Better is fixing the broken homeownership system by combining our proprietary technology with a seamless customer service experience while removing the middleman and saving customers thousands in unnecessary charges and fees

Better

better

SaaS is for Suckers: AI Already Builds Better Custom Software

Can WebFlux Be Better Than Spring MVC? What You Need to Know

Building applications that focus on tackling complex business problems and at the same time ensuring scalability, performance, and user-centricity may prove to be tricky. Artificial intelligence (AI) is indeed the technology that brings this change, which enables teams to put out data-driven insights plus fast-paced innovation. Our article will describe how AI-powered insights bring about revolutionary changes in product engineering; what technology has made such advances possible; and other lessons learned through actual cases showing the force of change. The Role of AI in Modern Product Engineering AI’s role in product engineering extends beyond automation. It involves extracting actionable insights from vast data repositories to enhance decision-making, streamline processes, and enable predictive capabilities. Let’s explore the core aspects where AI is making a difference: 1. Enhanced Requirements Gathering and Analysis Traditional requirements gathering can be time-intensive and error-prone. AI-powered tools like natural language processing (NLP) and sentiment analysis enable automatic parsing of customer feedback, market trends, and stakeholder input. These insights lead to accurate requirement specifications and faster product ideation cycles. Example: AI models can analyze customer reviews and feedback to identify recurring pain points, which can then guide feature prioritization during application development. 2. AI in Design Optimization AI plays a pivotal role in generating optimized designs by leveraging generative design algorithms. By analyzing design parameters, constraints, and objectives, these algorithms create multiple design variants, allowing engineers to select the most efficient and cost-effective options. Technology in Use: Generative Adversarial Networks (GANs): These models simulate design variations and enable testing for performance improvements.
AI-Driven UI/UX Optimization: AI-powered heatmaps and user interaction analytics ensure intuitive and user-friendly application interfaces. 3. Predictive Development and Proactive Maintenance AI-driven predictive analytics facilitates the identification of potential issues before they manifest. Machine learning models trained on historical application data can forecast performance bottlenecks, hardware failures, or security vulnerabilities. Key Features: Error Prediction Models: Tools like TensorFlow and PyTorch enable the development of models that identify code-level vulnerabilities.
Automated Code Refactoring: AI-powered tools like DeepCode and Tabnine assist in suggesting optimal code refactoring strategies. 4. Automated Testing and Quality Assurance AI automates extensive testing cycles by identifying test cases, generating test scripts, and even executing tests autonomously. This dramatically reduces testing time while ensuring thorough coverage. Technologies Driving Automation: AI in Functional Testing: Tools like Test.ai use machine learning to simulate user behaviors and validate application functionalities.
Performance Testing with AI: AI-driven tools like Apache JMeter integrated with predictive algorithms can simulate large-scale workloads to test application resilience. 5. AI-Driven Insights for Scalability Scalability is a cornerstone of modern product engineering. AI-powered analytics assess current system usage patterns and project future demands. This ensures proactive resource scaling, avoiding downtime or performance degradation. Implementation Example: Cloud platforms like AWS and Azure leverage AI to optimize resource allocation, enabling applications to scale seamlessly based on real-time demand patterns. Enabling Technologies for AI-Driven Product Engineering AI-driven insights in product engineering are powered by a suite of advanced technologies. Let’s look at the tools and frameworks enabling these capabilities: 1. Machine Learning and Deep Learning Machine learning (ML) models identify patterns and trends in data, which can be applied to optimize development workflows, predict system performance, and detect anomalies. Popular Frameworks: TensorFlow: For building predictive and classification models.
Scikit-learn: For statistical modeling and data mining. 2. Natural Language Processing NLP processes textual data, enabling tools to parse requirements, user feedback, and documentation. It also powers AI-driven documentation assistants. Technologies: SpaCy: For linguistic analysis.
BERT (Bidirectional Encoder Representations from Transformers): For contextual understanding in large text datasets. 3. Cloud Computing and AI Services Cloud providers offer integrated AI services that streamline application engineering. These services include automated data analytics pipelines, serverless architectures, and pre-trained models. Example Providers: AWS AI Services: For model training and deployment.
Google AI: For BigQuery ML integration. Benefits of AI-Driven Insights in Product Engineering 1. Accelerated Time-to-Market AI optimizes development cycles through automation, predictive modeling, and intelligent decision-making, enabling faster delivery. 2. Improved Application Quality With AI-driven testing and predictive maintenance, applications achieve higher performance, reliability, and user satisfaction. 3. Cost Efficiency AI optimizes resource allocation, reduces manual efforts, and minimizes rework, resulting in cost savings throughout the engineering lifecycle. 4. Enhanced Innovation AI fosters innovation by uncovering latent insights in data, inspiring novel solutions and features. Challenges and Mitigation Strategies Despite its advantages, adopting AI in product engineering comes with challenges: 1. Data Privacy and Security AI relies on extensive data collection, raising concerns about data privacy and compliance with regulations like GDPR. Solution: Implement robust data anonymization and encryption techniques. 2. Skill Gaps AI integration requires skilled personnel familiar with machine learning, data science, and software engineering. Solution: Invest in employee upskilling programs and leverage user-friendly AI platforms. 3. Integration Complexity Integrating AI into legacy systems can be complex and resource-intensive. Solution: Adopt modular AI solutions and scalable cloud-based AI frameworks. How Indium Tech Helps Accelerate Innovation in Product Engineering Indium stands at the forefront of AI-driven product engineering, offering end-to-end services tailored to the unique needs of enterprises. With a focus on innovation and operational excellence, Indium’s solutions empower businesses to achieve scalability, optimize performance, and deliver exceptional user experiences. Key Offerings: AI-Powered Application Development: Custom built AI methods integrated throughout the application lifecycle.
Comprehensive Testing Services: The use of AI to ensure quality releases through the automation of testing.
Data Analytics Expertise: Open up value insight from the enterprise data to stimulate innovations.
Cloud & DevOps Integration: AI-driven CI/CD pipelines for rigorous deployment practices. Conclusion AI-powered insights transform the limits of product engineering with new concepts about product ideation, design, and delivery. Through predictive capabilities, automation, and data-driven insights, organizations can accelerate innovation, optimize operations, and enter emerging markets more quickly than their competitors. Get in touch with Indium’s experts to lead your organization through this transformation by delivering custom product engineering solutions. Building applications that focus on tackling complex business problems and at the same time ensuring scalability, performance, and user-centricity may prove to be tricky. Artificial intelligence (AI) is indeed the technology that brings this change, which enables teams to put out data-driven insights plus fast-paced innovation. Our article will describe how AI-powered insights bring about revolutionary changes in product engineering; what technology has made such advances possible; and other lessons learned through actual cases showing the force of change. The Role of AI in Modern Product Engineering AI’s role in product engineering extends beyond automation. It involves extracting actionable insights from vast data repositories to enhance decision-making, streamline processes, and enable predictive capabilities. Let’s explore the core aspects where AI is making a difference: AI’s role in product engineering 1. Enhanced Requirements Gathering and Analysis 1. Enhanced Requirements Gathering and Analysis Traditional requirements gathering can be time-intensive and error-prone. AI-powered tools like natural language processing (NLP) and sentiment analysis enable automatic parsing of customer feedback, market trends, and stakeholder input. These insights lead to accurate requirement specifications and faster product ideation cycles. natural language processing Example: AI models can analyze customer reviews and feedback to identify recurring pain points, which can then guide feature prioritization during application development. Example: 2. AI in Design Optimization 2. AI in Design Optimization AI plays a pivotal role in generating optimized designs by leveraging generative design algorithms. By analyzing design parameters, constraints, and objectives, these algorithms create multiple design variants, allowing engineers to select the most efficient and cost-effective options. Technology in Use: Technology in Use: Generative Adversarial Networks (GANs): These models simulate design variations and enable testing for performance improvements. AI-Driven UI/UX Optimization: AI-powered heatmaps and user interaction analytics ensure intuitive and user-friendly application interfaces. Generative Adversarial Networks (GANs): These models simulate design variations and enable testing for performance improvements. Generative Adversarial Networks (GANs): AI-Driven UI/UX Optimization: AI-powered heatmaps and user interaction analytics ensure intuitive and user-friendly application interfaces. AI-Driven UI/UX Optimization: 3. Predictive Development and Proactive Maintenance 3. Predictive Development and Proactive Maintenance AI-driven predictive analytics facilitates the identification of potential issues before they manifest. Machine learning models trained on historical application data can forecast performance bottlenecks, hardware failures, or security vulnerabilities. Key Features: Key Features: Error Prediction Models: Tools like TensorFlow and PyTorch enable the development of models that identify code-level vulnerabilities. Automated Code Refactoring: AI-powered tools like DeepCode and Tabnine assist in suggesting optimal code refactoring strategies. Error Prediction Models: Tools like TensorFlow and PyTorch enable the development of models that identify code-level vulnerabilities. Error Prediction Models: Automated Code Refactoring: AI-powered tools like DeepCode and Tabnine assist in suggesting optimal code refactoring strategies. Automated Code Refactoring: 4. Automated Testing and Quality Assurance 4. Automated Testing and Quality Assurance AI automates extensive testing cycles by identifying test cases, generating test scripts, and even executing tests autonomously. This dramatically reduces testing time while ensuring thorough coverage. Technologies Driving Automation: Technologies Driving Automation: AI in Functional Testing: Tools like Test.ai use machine learning to simulate user behaviors and validate application functionalities. Performance Testing with AI: AI-driven tools like Apache JMeter integrated with predictive algorithms can simulate large-scale workloads to test application resilience. AI in Functional Testing: Tools like Test.ai use machine learning to simulate user behaviors and validate application functionalities. AI in Functional Testing: Performance Testing with AI: AI-driven tools like Apache JMeter integrated with predictive algorithms can simulate large-scale workloads to test application resilience. Performance Testing with AI: 5. AI-Driven Insights for Scalability 5. AI-Driven Insights for Scalability Scalability is a cornerstone of modern product engineering. AI-powered analytics assess current system usage patterns and project future demands. This ensures proactive resource scaling, avoiding downtime or performance degradation. Implementation Example: Implementation Example: Cloud platforms like AWS and Azure leverage AI to optimize resource allocation, enabling applications to scale seamlessly based on real-time demand patterns. Cloud platforms like AWS and Azure leverage AI to optimize resource allocation, enabling applications to scale seamlessly based on real-time demand patterns. Enabling Technologies for AI-Driven Product Engineering AI-driven insights in product engineering are powered by a suite of advanced technologies. Let’s look at the tools and frameworks enabling these capabilities: 1. Machine Learning and Deep Learning 1. Machine Learning and Deep Learning Machine learning (ML) models identify patterns and trends in data, which can be applied to optimize development workflows, predict system performance, and detect anomalies. Popular Frameworks: Popular Frameworks: TensorFlow: For building predictive and classification models. Scikit-learn: For statistical modeling and data mining. TensorFlow: For building predictive and classification models. TensorFlow: Scikit-learn: For statistical modeling and data mining. Scikit-learn: 2. Natural Language Processing 2. Natural Language Processing NLP processes textual data, enabling tools to parse requirements, user feedback, and documentation. It also powers AI-driven documentation assistants. Technologies: Technologies: SpaCy: For linguistic analysis. BERT (Bidirectional Encoder Representations from Transformers): For contextual understanding in large text datasets. SpaCy: For linguistic analysis. SpaCy: BERT (Bidirectional Encoder Representations from Transformers): For contextual understanding in large text datasets. BERT (Bidirectional Encoder Representations from Transformers): 3. Cloud Computing and AI Services 3. Cloud Computing and AI Services Cloud providers offer integrated AI services that streamline application engineering. These services include automated data analytics pipelines, serverless architectures, and pre-trained models. AI services AI services Example Providers: Example Providers: AWS AI Services: For model training and deployment. Google AI: For BigQuery ML integration. AWS AI Services: For model training and deployment. AWS AI Services: Google AI: For BigQuery ML integration. Google AI: Benefits of AI-Driven Insights in Product Engineering 1. Accelerated Time-to-Market 1. Accelerated Time-to-Market AI optimizes development cycles through automation, predictive modeling, and intelligent decision-making, enabling faster delivery. 2. Improved Application Quality 2. Improved Application Quality With AI-driven testing and predictive maintenance, applications achieve higher performance, reliability, and user satisfaction. 3. Cost Efficiency 3. Cost Efficiency AI optimizes resource allocation, reduces manual efforts, and minimizes rework, resulting in cost savings throughout the engineering lifecycle. 4. Enhanced Innovation 4. Enhanced Innovation AI fosters innovation by uncovering latent insights in data, inspiring novel solutions and features. Challenges and Mitigation Strategies Despite its advantages, adopting AI in product engineering comes with challenges: 1. Data Privacy and Security 1. Data Privacy and Security AI relies on extensive data collection, raising concerns about data privacy and compliance with regulations like GDPR. Solution: Implement robust data anonymization and encryption techniques. Solution: 2. Skill Gaps 2. Skill Gaps AI integration requires skilled personnel familiar with machine learning, data science, and software engineering. Solution: Invest in employee upskilling programs and leverage user-friendly AI platforms. Solution: 3. Integration Complexity 3. Integration Complexity Integrating AI into legacy systems can be complex and resource-intensive. Solution: Adopt modular AI solutions and scalable cloud-based AI frameworks. Solution: How Indium Tech Helps Accelerate Innovation in Product Engineering Indium stands at the forefront of AI-driven product engineering, offering end-to-end services tailored to the unique needs of enterprises. With a focus on innovation and operational excellence, Indium’s solutions empower businesses to achieve scalability, optimize performance, and deliver exceptional user experiences. Key Offerings: Key Offerings: AI-Powered Application Development: Custom built AI methods integrated throughout the application lifecycle. Comprehensive Testing Services: The use of AI to ensure quality releases through the automation of testing. Data Analytics Expertise: Open up value insight from the enterprise data to stimulate innovations. Cloud & DevOps Integration: AI-driven CI/CD pipelines for rigorous deployment practices. AI-Powered Application Development: Custom built AI methods integrated throughout the application lifecycle. AI-Powered Application Development: Comprehensive Testing Services: The use of AI to ensure quality releases through the automation of testing. Comprehensive Testing Services: Data Analytics Expertise: Open up value insight from the enterprise data to stimulate innovations. Data Analytics Expertise: Cloud & DevOps Integration: AI-driven CI/CD pipelines for rigorous deployment practices. Cloud & DevOps Integration: Conclusion AI-powered insights transform the limits of product engineering with new concepts about product ideation, design, and delivery. Through predictive capabilities, automation, and data-driven insights, organizations can accelerate innovation, optimize operations, and enter emerging markets more quickly than their competitors. Get in touch with Indium’s experts to lead your organization through this transformation by delivering custom product engineering solutions.

AI Is Making it Easier to Engineer Better Products—Here's How

AI ukax juk’amp suma yänak ingeniero lurañax juk’amp jasakiw —Akanx Kunjams lurasi

AI gør det nemmere at konstruere bedre produkter – her er hvordan

KI erleichtert die Entwicklung besserer Produkte – so geht‘s

La IA facilita la creación de mejores productos: aquí te contamos cómo

Az AI megkönnyíti a jobb termékek tervezését – íme, hogyan

AI מקל על הנדסת מוצרים טובים יותר - הנה איך

AI によってより優れた製品の開発が容易になる — その方法とは

AI ले अझ राम्रो उत्पादनहरू इन्जिनियर गर्न सजिलो बनाउँदैछ—यहाँ कसरी छ

AI maakt het makkelijker om betere producten te ontwikkelen - en wel zo

AI gör det enklare att konstruera bättre produkter – så här

AI inereneriň has gowy önümlerini has aňsatlaşdyrýar - Ine

AI بہتر پروڈکٹس کو انجینئر کرنا آسان بنا رہا ہے — یہ طریقہ ہے۔

Are you ready to level up your skills as a developer in 2025? In this article, I show 10 books that will help you become a better programmer, enhance your coding practices, and improve your overall software engineering skills. Whether you’re a beginner or an experienced developer, these books cover essential topics like clean coding, debugging, system design, productivity, and more. My selection is based on suggestions and recommendations of top developers worldwide. If you are ready, then let’s dive in. 1. Clean Code: A Handbook of Agile Software Craftsmanship by Robert C. Martin I still believe this book must be in your collection, even if it is a bit old. This book emphasizes the importance of writing code that is not only functional but also easy to read and maintain. It provides actionable principles, rules, and examples to help developers craft professional-grade software. From naming conventions to refactoring techniques, Clean Code is a goldmine of practical advice. It’s essential for improving code readability, maintainability, and professionalism. If you want to elevate your coding standards and write code that others can easily work with, this book is a must. 2. The Pragmatic Programmer: Your Journey to Mastery by Andrew Hunt and David Thomas The Pragmatic Programmer covers everything from debugging techniques to automation and career advice. It encourages critical thinking and problem-solving while providing actionable strategies to tackle real-world challenges in software development. This book teaches you how to think like a pragmatic developer — someone who focuses on efficiency, adaptability, and continuous improvement. 3. Code Complete: A Practical Handbook of Software Construction by Steve McConnell Often referred to as the “manual of software engineering,” Code Complete dives deep into the principles of writing high-quality code. It covers topics like design patterns, testing strategies, debugging techniques, and much more. If you want an all-encompassing resource on writing bug-free, maintainable code while understanding the entire software development lifecycle, this is it. 4. Why Programs Fail: A Guide to Systematic Debugging Debugging can be one of the most time-consuming aspects of programming. This book teaches systematic approaches to identifying and fixing bugs quickly without relying on trial-and-error methods like excessive console logging. Debugging is one of the most critical skills for any developer. This book will save you countless hours by teaching you how to debug efficiently. 5. Deep Work: Rules for Focused Success in a Distracted World by Cal Newport In Deep Work, Cal Newport explains how cultivating focus can lead to extraordinary results in both personal and professional life. He provides actionable strategies for minimizing distractions and maximizing productivity. As developers, our work often requires long periods of uninterrupted focus. This book will teach you how to protect your attention and get into “the zone.” 6. Refactoring: Improving the Design of Existing Code by Martin Fowler Martin Fowler’s Refactoring introduces the concept of systematically improving the design of existing code without altering its functionality. The book provides a catalog of over 70 refactoring techniques, each with step-by-step instructions, motivations, and examples. It also discusses how to recognize “code smells” that indicate opportunities for refactoring and emphasizes the importance of testing during the process. The second edition includes updated examples in JavaScript, making it accessible to modern developers. This book is essential for developers who want to enhance the readability, maintainability, and scalability of their codebases. It teaches how to refactor safely and efficiently while minimizing the risk of introducing bugs. 7. System Design Interview by Alex Xu This is my favorite book about system design. This book breaks down complex system design concepts into digestible chunks, making it easier for developers to understand distributed systems, scalability, and performance optimization — key topics in technical interviews at top companies. If you’re preparing for system design interviews or want to deepen your knowledge of scalable architecture, this book is invaluable. 8. Don’t Make Me Think: A Common Sense Approach to Web Usability by Steve Krug This book focuses on creating web interfaces that are simple and user-friendly. It emphasizes usability testing and offers practical advice on designing websites that users can navigate effortlessly without confusion or frustration. For web developers aiming to improve user experience (UX), this is an essential read that simplifies complex usability concepts into actionable advice. 9. Design Patterns: Elements of Reusable Object-Oriented Software by Erich Gamma This classic introduces 23 design patterns that solve common software design problems. It explains when and how to use these patterns effectively while improving code reusability and scalability. Understanding design patterns is crucial for building robust object-oriented systems that are easy to maintain and extend over time. 10. Domain-Driven Design: Tackling Complexity in the Heart of Software by Eric Evans This book introduces the concept of aligning your software model closely with business domains. It provides strategies for tackling complexity using DDD principles like ubiquitous language and bounded contexts. If you’re working on complex systems or collaborating closely with business stakeholders, this book will teach you how to bridge the gap between technical solutions and business needs effectively. 10. Domain-Driven Design: Tackling Complexity in the Heart of Software by Eric Evans This book introduces the concept of aligning your software model closely with business domains. It provides strategies for tackling complexity using DDD principles like ubiquitous language and bounded contexts. Domain-Driven Design: Tackling Complexity in the Heart of Software by Eric Evans If you’re working on complex systems or collaborating closely with business stakeholders, this book will teach you how to bridge the gap between technical solutions and business needs effectively. Video :-) https://www.youtube.com/watch?v=F0fVOnfDZtM&embedable=true Conclusion Set the goal to read these books in 2025, and your programming skills will never be the same. ;) If you want to share your favorite books, please do so in the comment below. If you want to buy the book, you can search in Google or use my link where I collect everything together: 10 Books to Read in 2025 to Become a Better Developer. P.S. This page contains affiliate links. If you make a purchase through these links, I may earn a commission at no additional cost to you. ;) Good luck! ;) Are you ready to level up your skills as a developer in 2025? In this article, I show 10 books that will help you become a better programmer, enhance your coding practices, and improve your overall software engineering skills. Whether you’re a beginner or an experienced developer, these books cover essential topics like clean coding, debugging, system design, productivity, and more. 10 books My selection is based on suggestions and recommendations of top developers worldwide. If you are ready, then let’s dive in. 1. Clean Code: A Handbook of Agile Software Craftsmanship by Robert C. Martin I still believe this book must be in your collection, even if it is a bit old. This book emphasizes the importance of writing code that is not only functional but also easy to read and maintain. It provides actionable principles, rules, and examples to help developers craft professional-grade software. From naming conventions to refactoring techniques, Clean Code is a goldmine of practical advice. Clean Code It’s essential for improving code readability, maintainability, and professionalism. If you want to elevate your coding standards and write code that others can easily work with, this book is a must. 2. The Pragmatic Programmer: Your Journey to Mastery by Andrew Hunt and David Thomas The Pragmatic Programmer covers everything from debugging techniques to automation and career advice. It encourages critical thinking and problem-solving while providing actionable strategies to tackle real-world challenges in software development. The Pragmatic Programmer This book teaches you how to think like a pragmatic developer — someone who focuses on efficiency, adaptability, and continuous improvement. 3. Code Complete: A Practical Handbook of Software Construction by Steve McConnell Often referred to as the “manual of software engineering,” Code Complete dives deep into the principles of writing high-quality code. It covers topics like design patterns, testing strategies, debugging techniques, and much more. Code Complete If you want an all-encompassing resource on writing bug-free, maintainable code while understanding the entire software development lifecycle, this is it. 4. Why Programs Fail: A Guide to Systematic Debugging Debugging can be one of the most time-consuming aspects of programming. This book teaches systematic approaches to identifying and fixing bugs quickly without relying on trial-and-error methods like excessive console logging. Debugging is one of the most critical skills for any developer. This book will save you countless hours by teaching you how to debug efficiently. 5. Deep Work: Rules for Focused Success in a Distracted World by Cal Newport In Deep Work , Cal Newport explains how cultivating focus can lead to extraordinary results in both personal and professional life. He provides actionable strategies for minimizing distractions and maximizing productivity. Deep Work As developers, our work often requires long periods of uninterrupted focus. This book will teach you how to protect your attention and get into “the zone.” 6. Refactoring: Improving the Design of Existing Code by Martin Fowler Martin Fowler’s Refactoring introduces the concept of systematically improving the design of existing code without altering its functionality. The book provides a catalog of over 70 refactoring techniques, each with step-by-step instructions, motivations, and examples. It also discusses how to recognize “code smells” that indicate opportunities for refactoring and emphasizes the importance of testing during the process. The second edition includes updated examples in JavaScript, making it accessible to modern developers. Refactoring This book is essential for developers who want to enhance the readability, maintainability, and scalability of their codebases. It teaches how to refactor safely and efficiently while minimizing the risk of introducing bugs. 7. System Design Interview by Alex Xu This is my favorite book about system design. This book breaks down complex system design concepts into digestible chunks, making it easier for developers to understand distributed systems, scalability, and performance optimization — key topics in technical interviews at top companies. If you’re preparing for system design interviews or want to deepen your knowledge of scalable architecture, this book is invaluable. 8. Don’t Make Me Think: A Common Sense Approach to Web Usability by Steve Krug This book focuses on creating web interfaces that are simple and user-friendly. It emphasizes usability testing and offers practical advice on designing websites that users can navigate effortlessly without confusion or frustration. For web developers aiming to improve user experience (UX), this is an essential read that simplifies complex usability concepts into actionable advice. 9. Design Patterns: Elements of Reusable Object-Oriented Software by Erich Gamma This classic introduces 23 design patterns that solve common software design problems. It explains when and how to use these patterns effectively while improving code reusability and scalability. Understanding design patterns is crucial for building robust object-oriented systems that are easy to maintain and extend over time. 10. Domain-Driven Design: Tackling Complexity in the Heart of Software by Eric Evans This book introduces the concept of aligning your software model closely with business domains. It provides strategies for tackling complexity using DDD principles like ubiquitous language and bounded contexts. If you’re working on complex systems or collaborating closely with business stakeholders, this book will teach you how to bridge the gap between technical solutions and business needs effectively. 10. Domain-Driven Design: Tackling Complexity in the Heart of Software by Eric Evans This book introduces the concept of aligning your software model closely with business domains. It provides strategies for tackling complexity using DDD principles like ubiquitous language and bounded contexts. Domain-Driven Design: Tackling Complexity in the Heart of Software by Eric Evans If you’re working on complex systems or collaborating closely with business stakeholders, this book will teach you how to bridge the gap between technical solutions and business needs effectively. Video :-) https://www.youtube.com/watch?v=F0fVOnfDZtM&embedable=true https://www.youtube.com/watch?v=F0fVOnfDZtM&embedable=true Conclusion Set the goal to read these books in 2025, and your programming skills will never be the same. ;) If you want to share your favorite books, please do so in the comment below. If you want to buy the book, you can search in Google or use my link where I collect everything together: 10 Books to Read in 2025 to Become a Better Developer . 10 Books to Read in 2025 to Become a Better Developer 10 Books to Read in 2025 to Become a Better Developer P.S. This page contains affiliate links. If you make a purchase through these links, I may earn a commission at no additional cost to you. ;) P.S. This page contains affiliate links. If you make a purchase through these links, I may earn a commission at no additional cost to you. ;) Good luck! ;)

Top 10 Software Development Books to Become a Better Software Developer in 2025

በ2025 የተሻሉ የሶፍትዌር ገንቢ ለመሆን 10 ምርጥ የሶፍትዌር ልማት መጽሐፍት።

Los 10 mejores libros de desarrollo de software para convertirse en un mejor desarrollador de software en 2025

10 parasta ohjelmistokehityskirjaa, joista tulee parempi ohjelmistokehittäjä vuonna 2025

Os 10 mellores libros de desenvolvemento de software para converterse nun mellor programador de software en 2025

2025 में बेहतर सॉफ्टवेयर डेवलपर बनने के लिए शीर्ष 10 सॉफ्टवेयर डेवलपमेंट पुस्तकें

2025 年に優れたソフトウェア開発者になるためのソフトウェア開発に関するベスト 10 書籍

10 geriausių programinės įrangos kūrimo knygų, kurios padės tapti geresniu programinės įrangos kūrėju 2025 m

Top 10 softwareontwikkelingsboeken om in 2025 een betere softwareontwikkelaar te worden

Epamusoro gumi Mabhuku Ekuvandudza Software Kuti Ave Zvirinani Software Developer muna 2025

2025 இல் சிறந்த மென்பொருள் உருவாக்குநராக மாறுவதற்கான சிறந்த 10 மென்பொருள் மேம்பாட்டு புத்தகங்கள்

ኣብ 2025 ዝሓሸ ሶፍትዌር ዲቨሎፐር ንምዃን ዝሕግዙ ብሉጻት 10 መጻሕፍቲ ምዕባለ ሶፍትዌር

Iincwadi ezili-10 eziphezulu zoPhuhliso lweSoftware ukuze ube nguMphuhlisi weSoftware oNgcono ngo-2025

This New AI Tool Claims to Solve Data Problems Better Than Anything Else—Here’s Why That Matters

Bu yeni AI aləti məlumat problemlərini hər şeydən daha yaxşı həll etdiyini iddia edir - bu, niyə vacibdir

Dieses neue KI-Tool soll Datenprobleme besser lösen als alles andere – deshalb ist das wichtig

Esta nueva herramienta de inteligencia artificial afirma resolver problemas de datos mejor que cualquier otra cosa: esta es la razón por la que es importante

AI tresna berri honek datu-arazoak beste ezer baino hobeto konpontzen dituela dio - Hona hemen zergatik garrantzitsua den

Alat AI Baru Ini Mengklaim Dapat Memecahkan Masalah Data Lebih Baik Daripada Apa Pun—Inilah Alasannya

この新しい AI ツールは、他の何よりも優れたデータ問題の解決を主張しています。これが重要な理由です。

ეს ახალი ხელოვნური ინტელექტის ინსტრუმენტი ირწმუნება, რომ მონაცემთა პრობლემებს სხვაზე უკეთ წყვეტს - აი, რატომ არის ეს მნიშვნელოვანი

Бұл жаңа AI құралы деректер мәселелерін кез келген нәрседен жақсырақ шешеді деп мәлімдейді - бұл неге маңызды

මෙම නව AI මෙවලම වෙනත් ඕනෑම දෙයකට වඩා හොඳින් දත්ත ගැටළු විසඳීමට හිමිකම් කියයි - මෙන්න එය වැදගත් වන්නේ ඇයි

Ichi Chitsva cheAI Tool Chinoti Kugadzirisa Matambudziko eData zvirinani pane chero chimwe chinhu-Hechi ndicho chikonzero nei zvichikosha.

Det här nya AI-verktyget hävdar att det löser dataproblem bättre än något annat - här är varför det är viktigt

Ushbu yangi AI vositasi ma'lumotlar muammolarini hamma narsadan yaxshiroq hal qilishni da'vo qilmoqda - bu nima uchun muhim

Authors:
(1) Martyna Wiącek, Institute of Computer Science, Polish Academy of Sciences;
(2) Piotr Rybak, Institute of Computer Science, Polish Academy of Sciences;
(3) Łukasz Pszenny, Institute of Computer Science, Polish Academy of Sciences;
(4) Alina Wróblewska, Institute of Computer Science, Polish Academy of Sciences. Editor's note: This is Part 5 of 10 of a study on improving the evaluation and comparison of tools used in natural language preprocessing. Read the rest below. Table of Links Abstract and 1. Introduction and related works NLPre benchmarking 2.1. Research concept 2.2. Online benchmarking system 2.3. Configuration NLPre-PL benchmark 3.1. Datasets 3.2. Tasks Evaluation 4.1. Evaluation methodology 4.2. Evaluated systems 4.3. Results Conclusions

Appendices
Acknowledgements
Bibliographical References
Language Resource References 3. NLPre-PL benchmark 3.1. Datasets NKJP1M (Przepiórkowski et al., 2018) The NKJP1M subcorpus of the Polish National Corpus (Przepiórkowski et al., 2012) is manually annotated according to the NKJP tagset (Szałkiewicz and Przepiórkowski, 2012) and afterwards modified in line with the Morfeusz tagset (Woliński, 2019). This balanced subset of thematic- and genre-diverse texts and transcriptions is used to train Polish POS taggers. NKJP1M is maintained in two formats: TEI[8] and DAG.[9] These two formats are accepted by older NLPre tools but not modern ones. We thus convert NKJP1M to the CoNLL-X format (Buchholz and Marsi, 2006) preserving the original segmentation, POS tags and morphological features (i.e. the Morfeusz tagset), and to the CoNLL-U format10 with UD tags, Morfeusz tags (XPOS) and UD morphological features. Since there is no generally accepted split of NKJP1M into training, development and testing subsets, we uniformly divide NKJP1M in all formats (i.e. DAG, TEI, CoNLL-X and CoNLL-U) pursuant to the formulated splitting heuristics. Each document in the subcorpus contains multiple paragraphs of continuous textual data. To avoid possible information leakage, we treat each such paragraph as an indivisible unit. To ensure that the subsets include paragraphs of varying length, we investigate the distribution over the number of segments in each paragraph. Since it is akin to Gaussian distribution, we decide to not exclude any data, and we divide the paragraphs into K = 10 buckets of roughly similar size and then sample from them with respective ratios of 0.8:0.1:0.1 (corresponding to train, dev, and test subsets). This data selection technique assures similar distribution of segments number per paragraph in three subsets, hereafter byName. For creating our second split, hereafter byType, we consider the type of document a paragraph belongs to. We first group paragraphs into categories equal to the document types, and then we repeat the above-mentioned procedure per category (see the summary of NKJP1M and data splits in Table 1). PDB-UD (Wróblewska, 2018) Polish Dependency Bank is the largest collection of Polish sentences manually annotated with dependency trees and afterwards converted into UD representations in line with the UD annotation schema (de Marneffe et al., 2021). PDB-UD slightly correlates with NKJP1M, i.e., a subset of the PDB-UD sentences comes from NKJP1M, and the language-specific tags (XPOS) in PDB-UD match the Morfeusz tagset. PDB-UD is typically used to train NLPre systems for Polish. In NLPre-PL, we use the original PDB-UD data without any modifications and its standard split (see the statistical summary of PDB-UD in Table 1). This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. [8] http://nlp.ipipan.waw.pl/TEI4NKJP. [9] https://github.com/kawu/concraft-pl#data-format [10] https://universaldependencies.org/format.html Authors: (1) Martyna Wiącek, Institute of Computer Science, Polish Academy of Sciences; (2) Piotr Rybak, Institute of Computer Science, Polish Academy of Sciences; (3) Łukasz Pszenny, Institute of Computer Science, Polish Academy of Sciences; (4) Alina Wróblewska, Institute of Computer Science, Polish Academy of Sciences. Authors: Authors: (1) Martyna Wiącek, Institute of Computer Science, Polish Academy of Sciences; (2) Piotr Rybak, Institute of Computer Science, Polish Academy of Sciences; (3) Łukasz Pszenny, Institute of Computer Science, Polish Academy of Sciences; (4) Alina Wróblewska, Institute of Computer Science, Polish Academy of Sciences. Editor's note: This is Part 5 of 10 of a study on improving the evaluation and comparison of tools used in natural language preprocessing. Read the rest below. Editor's note: This is Part 5 of 10 of a study on improving the evaluation and comparison of tools used in natural language preprocessing. Read the rest below. Editor's note: This is Part 5 of 10 of a study on improving the evaluation and comparison of tools used in natural language preprocessing. Read the rest below. Editor's note: This is Part 5 of 10 of a study on improving the evaluation and comparison of tools used in natural language preprocessing. Read the rest below. Table of Links Abstract and 1. Introduction and related works Abstract and 1. Introduction and related works NLPre benchmarking NLPre benchmarking 2.1. Research concept 2.1. Research concept 2.2. Online benchmarking system 2.2. Online benchmarking system 2.3. Configuration 2.3. Configuration NLPre-PL benchmark NLPre-PL benchmark 3.1. Datasets 3.1. Datasets 3.2. Tasks 3.2. Tasks Evaluation Evaluation 4.1. Evaluation methodology 4.1. Evaluation methodology 4.2. Evaluated systems 4.2. Evaluated systems 4.3. Results 4.3. Results Conclusions

Appendices
Acknowledgements
Bibliographical References
Language Resource References Conclusions Appendices
Acknowledgements
Bibliographical References
Language Resource References Conclusions Appendices Acknowledgements Bibliographical References Language Resource References Appendices Acknowledgements Bibliographical References Language Resource References 3. NLPre-PL benchmark 3.1. Datasets NKJP1M (Przepiórkowski et al., 2018) The NKJP1M subcorpus of the Polish National Corpus (Przepiórkowski et al., 2012) is manually annotated according to the NKJP tagset (Szałkiewicz and Przepiórkowski, 2012) and afterwards modified in line with the Morfeusz tagset (Woliński, 2019). This balanced subset of thematic- and genre-diverse texts and transcriptions is used to train Polish POS taggers. NKJP1M is maintained in two formats: TEI[8] and DAG.[9] These two formats are accepted by older NLPre tools but not modern ones. We thus convert NKJP1M to the CoNLL-X format (Buchholz and Marsi, 2006) preserving the original segmentation, POS tags and morphological features (i.e. the Morfeusz tagset), and to the CoNLL-U format10 with UD tags, Morfeusz tags (XPOS) and UD morphological features. NKJP1M Since there is no generally accepted split of NKJP1M into training, development and testing subsets, we uniformly divide NKJP1M in all formats (i.e. DAG, TEI, CoNLL-X and CoNLL-U) pursuant to the formulated splitting heuristics. Each document in the subcorpus contains multiple paragraphs of continuous textual data. To avoid possible information leakage, we treat each such paragraph as an indivisible unit. To ensure that the subsets include paragraphs of varying length, we investigate the distribution over the number of segments in each paragraph. Since it is akin to Gaussian distribution, we decide to not exclude any data, and we divide the paragraphs into K = 10 buckets of roughly similar size and then sample from them with respective ratios of 0.8:0.1:0.1 (corresponding to train, dev, and test subsets). This data selection technique assures similar distribution of segments number per paragraph in three subsets, hereafter byName. For creating our second split, hereafter byType, we consider the type of document a paragraph belongs to. We first group paragraphs into categories equal to the document types, and then we repeat the above-mentioned procedure per category (see the summary of NKJP1M and data splits in Table 1). PDB-UD (Wróblewska, 2018) Polish Dependency Bank is the largest collection of Polish sentences manually annotated with dependency trees and afterwards converted into UD representations in line with the UD annotation schema (de Marneffe et al., 2021). PDB-UD slightly correlates with NKJP1M, i.e., a subset of the PDB-UD sentences comes from NKJP1M, and the language-specific tags (XPOS) in PDB-UD match the Morfeusz tagset. PDB-UD is typically used to train NLPre systems for Polish. In NLPre-PL, we use the original PDB-UD data without any modifications and its standard split (see the statistical summary of PDB-UD in Table 1). PDB-UD This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. This paper is available on arxiv under CC BY-NC-SA 4.0 DEED license. available on arxiv [8] http://nlp.ipipan.waw.pl/TEI4NKJP. [9] https://github.com/kawu/concraft-pl#data-format [10] https://universaldependencies.org/format.html

New Framework Promises to Train AI to Better Understand Hard-to-Grasp Languages Like Polish

Table of Links Abstract and 1. Introduction 2. Preliminaries and 2.1. Blind deconvolution 2.2. Quadratic neural networks 3. Methodology 3.1. Time domain quadratic convolutional filter 3.2. Superiority of cyclic features extraction by QCNN 3.3. Frequency domain linear filter with envelope spectrum objective function 3.4. Integral optimization with uncertainty-aware weighing scheme 4. Computational experiments 4.1. Experimental configurations 4.2. Case study 1: PU dataset 4.3. Case study 2: JNU dataset 4.4. Case study 3: HIT dataset 5. Computational experiments 5.1. Comparison of BD methods 5.2. Classification results on various noise conditions 5.3. Employing ClassBD to deep learning classifiers 5.4. Employing ClassBD to machine learning classifiers 5.5. Feature extraction ability of quadratic and conventional networks 5.6. Comparison of ClassBD filters 6. Conclusions Appendix and References 5.3. Employing ClassBD to deep learning classifiers Given that ClassBD serves as a signal preprocessing module, it possesses the flexibility to be integrated into various backbone networks. In the context of this experiment, we assess the classification performance by incorporating ClassBD into four widely recognized deep learning classifiers: ResNet [85], MobileNetV3 [86], WDCNN [54], and Transformer [87]. It is important to note that some networks were initially designed for image classification, so some parameters are revised to accommodate the 1D signal input. The properties of these backbone networks are illustrated in Table 11. Firstly, the classification results on the PU dataset are illustrated in Table 12. Predominantly, ClassBD has demonstrated its efficacy in enhancing performance. For instance, in the case of N09M07F10 with -4dB noise, the application of ClassBD results in an improvement exceeding 10% across all backbones. On average, when the SNR is -4dB, ClassBD achieves F1 scores of 87.60%, whereas the raw networks only yield 64.07%. Subsequently, we also test these models to the JNU and HIT datasets, setting the SNRs to -6dB and -4dB respectively. As depicted in Table 13, ClassBD exhibits commendable performance on both datasets, and all backbones experience a performance boost. Specifically, on the JNU dataset, the F1 scores exceed 90% post the employment of ClassBD, irrespective of the backbone performance. Furthermore, a substantial improvement is observed in the Transformer. The raw Transformer initially yields F1 scores around 50% on the JNU and HIT datasets, which, after the application of ClassBD, escalates to an average F1 score of 90%. Finally, the results on the three datasets suggest that ClassBD can function as a plug-and-play denoising module, thereby enhancing the performance of deep learning classifiers under substantial noise. Considering both performance and model size, the simplest backbone, WDCNN, achieves consistent performance under all conditions. Consequently, we recommend it as the backbone for bearing fault diagnosis. Authors:
(1) Jing-Xiao Liao, Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China and School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China;
(2) Chao He, School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing, China;
(3) Jipu Li, Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China;
(4) Jinwei Sun, School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China;
(5) Shiping Zhang (Corresponding author), School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China;
(6) Xiaoge Zhang (Corresponding author), Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. Table of Links Abstract and 1. Introduction Abstract and 1. Introduction 2. Preliminaries and 2.1. Blind deconvolution 2. Preliminaries and 2.1. Blind deconvolution 2.2. Quadratic neural networks 2.2. Quadratic neural networks 3. Methodology 3. Methodology 3.1. Time domain quadratic convolutional filter 3.1. Time domain quadratic convolutional filter 3.2. Superiority of cyclic features extraction by QCNN 3.2. Superiority of cyclic features extraction by QCNN 3.3. Frequency domain linear filter with envelope spectrum objective function 3.3. Frequency domain linear filter with envelope spectrum objective function 3.4. Integral optimization with uncertainty-aware weighing scheme 3.4. Integral optimization with uncertainty-aware weighing scheme 4. Computational experiments 4.1. Experimental configurations 4.1. Experimental configurations 4.2. Case study 1: PU dataset 4.2. Case study 1: PU dataset 4.3. Case study 2: JNU dataset 4.3. Case study 2: JNU dataset 4.4. Case study 3: HIT dataset 4.4. Case study 3: HIT dataset 5. Computational experiments 5.1. Comparison of BD methods 5.1. Comparison of BD methods 5.2. Classification results on various noise conditions 5.2. Classification results on various noise conditions 5.3. Employing ClassBD to deep learning classifiers 5.3. Employing ClassBD to deep learning classifiers 5.4. Employing ClassBD to machine learning classifiers 5.4. Employing ClassBD to machine learning classifiers 5.5. Feature extraction ability of quadratic and conventional networks 5.5. Feature extraction ability of quadratic and conventional networks 5.6. Comparison of ClassBD filters 5.6. Comparison of ClassBD filters 6. Conclusions 6. Conclusions Appendix and References Appendix and References 5.3. Employing ClassBD to deep learning classifiers Given that ClassBD serves as a signal preprocessing module, it possesses the flexibility to be integrated into various backbone networks. In the context of this experiment, we assess the classification performance by incorporating ClassBD into four widely recognized deep learning classifiers: ResNet [85], MobileNetV3 [86], WDCNN [54], and Transformer [87]. It is important to note that some networks were initially designed for image classification, so some parameters are revised to accommodate the 1D signal input. The properties of these backbone networks are illustrated in Table 11. Firstly, the classification results on the PU dataset are illustrated in Table 12. Predominantly, ClassBD has demonstrated its efficacy in enhancing performance. For instance, in the case of N09M07F10 with -4dB noise, the application of ClassBD results in an improvement exceeding 10% across all backbones. On average, when the SNR is -4dB, ClassBD achieves F1 scores of 87.60%, whereas the raw networks only yield 64.07%. Subsequently, we also test these models to the JNU and HIT datasets, setting the SNRs to -6dB and -4dB respectively. As depicted in Table 13, ClassBD exhibits commendable performance on both datasets, and all backbones experience a performance boost. Specifically, on the JNU dataset, the F1 scores exceed 90% post the employment of ClassBD, irrespective of the backbone performance. Furthermore, a substantial improvement is observed in the Transformer. The raw Transformer initially yields F1 scores around 50% on the JNU and HIT datasets, which, after the application of ClassBD, escalates to an average F1 score of 90%. Finally, the results on the three datasets suggest that ClassBD can function as a plug-and-play denoising module, thereby enhancing the performance of deep learning classifiers under substantial noise. Considering both performance and model size, the simplest backbone, WDCNN, achieves consistent performance under all conditions. Consequently, we recommend it as the backbone for bearing fault diagnosis. Authors: (1) Jing-Xiao Liao, Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China and School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China; (2) Chao He, School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing, China; (3) Jipu Li, Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China; (4) Jinwei Sun, School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China; (5) Shiping Zhang (Corresponding author), School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China; (6) Xiaoge Zhang (Corresponding author), Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China. Authors: Authors: (1) Jing-Xiao Liao, Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China and School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China; (2) Chao He, School of Mechanical, Electronic and Control Engineering, Beijing Jiaotong University, Beijing, China; (3) Jipu Li, Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China; (4) Jinwei Sun, School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China; (5) Shiping Zhang (Corresponding author), School of Instrumentation Science and Engineering, Harbin Institute of Technology, Harbin, China; (6) Xiaoge Zhang (Corresponding author), Department of Industrial and Systems Engineering, The Hong Kong Polytechnic University, Hong Kong, Special Administrative Region of China. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. This paper is available on arxiv under CC by 4.0 Deed (Attribution 4.0 International) license. available on arxiv

ClassBD Boosts Deep Learning Classifiers for Better Fault Diagnosis

Authors:
(1) Shehan Munasinghe, Mohamed bin Zayed University of AI and Equal Contribution;
(2) Rusiru Thushara, Mohamed bin Zayed University of AI and Equal Contribution;
(3) Muhammad Maaz, Mohamed bin Zayed University of AI;
(4) Hanoona Abdul Rasheed, Mohamed bin Zayed University of AI;
(5) Salman Khan, Mohamed bin Zayed University of AI and Australian National University;
(6) Mubarak Shah, University of Central Florida;
(7) Fahad Khan, Mohamed bin Zayed University of AI and Linköping University. Editor's Note: This is Part 6 of 10 of a study detailing the development of a smarter AI model for videos. Read the rest below. Table of Links Abstract and 1 Introduction
2. Related Works


PG-Video-LLaVA


3.1. Overview
3.2. Architecture


Experiments


4.1. Implementation Details
4.2. Stronger Baseline
4.3. Spatial Grounding in Videos
4.4. Zero-Shot Visual Question Answering
5. Conclusion and References Supplementary Material A. Audio Modality Integration
B. Visual Grounding: Quantitative Evaluation
C. Qualitative Results for Visual Grounding
D. Quantitative Evaluations of Video-based Conversation Performance 4.2. Stronger Baseline This section provides an overview of the quantitative evaluations conducted to determine the effects of the strengthened baseline on PG-Video-LLaVA. To evaluate the impact of the enhanced baseline on PG-Video-LLaVA, we apply the benchmarking framework from Video-ChatGPT[22]. This framework measures performance on several axes critical for video-based conversational agents, including correctness of information, detail orientation, contextual understanding, temporal understanding, and consistency. In order to facilitate a reliable and reproducible evaluation, we have updated our assessment pipeline by replacing GPT-3.5-Turbo with Vicuna-13b-v1.5. This adjustment addresses the limitations in reproducibility inherent to the closed-source nature of GPT-3.5-Turbo. Subsequently, we have re-assessed both PG-Video-LLaVA and other recent models to ensure a fair and consistent comparison. The results shown in Table 1 demonstrate that PG-Video-LLaVA outperforms the foundational Video-ChatGPT model and exhibits superior performance when compared to other recent contributions in the domain. Following the quantitative assessment, the qualitative results in Figure 3 indicate the enhanced baseline’s impact on PG-Video-LLaVA’s performance. The PG-Video-LLaVA (13B) model exhibits improved accuracy in the information presented, a deeper level of descriptive detail, and a stronger alignment with the context and temporal progression of the videos. This advancement is particularly noticeable in the precise depiction of the child’s engagement with their surroundings and the giraffe’s behaviour, indicating a refined interpretation of both the activities and their settings. These qualitative insights are consistent with the quantitative results, highlighting the augmented baseline’s role in advancing PG-Video-LLaVA’s capacity in video understanding. This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Shehan Munasinghe, Mohamed bin Zayed University of AI and Equal Contribution; (2) Rusiru Thushara, Mohamed bin Zayed University of AI and Equal Contribution; (3) Muhammad Maaz, Mohamed bin Zayed University of AI; (4) Hanoona Abdul Rasheed, Mohamed bin Zayed University of AI; (5) Salman Khan, Mohamed bin Zayed University of AI and Australian National University; (6) Mubarak Shah, University of Central Florida; (7) Fahad Khan, Mohamed bin Zayed University of AI and Linköping University. Authors: Authors: (1) Shehan Munasinghe, Mohamed bin Zayed University of AI and Equal Contribution; (2) Rusiru Thushara, Mohamed bin Zayed University of AI and Equal Contribution; (3) Muhammad Maaz, Mohamed bin Zayed University of AI; (4) Hanoona Abdul Rasheed, Mohamed bin Zayed University of AI; (5) Salman Khan, Mohamed bin Zayed University of AI and Australian National University; (6) Mubarak Shah, University of Central Florida; (7) Fahad Khan, Mohamed bin Zayed University of AI and Linköping University. Editor's Note: This is Part 6 of 10 of a study detailing the development of a smarter AI model for videos. Read the rest below. Editor's Note: This is Part 6 of 10 of a study detailing the development of a smarter AI model for videos. Read the rest below. Editor's Note: This is Part 6 of 10 of a study detailing the development of a smarter AI model for videos. Read the rest below. Editor's Note: This is Part 6 of 10 of a study detailing the development of a smarter AI model for videos. Read the rest below. Table of Links Abstract and 1 Introduction 2. Related Works PG-Video-LLaVA 3.1. Overview 3.2. Architecture Experiments 4.1. Implementation Details 4.2. Stronger Baseline 4.3. Spatial Grounding in Videos 4.4. Zero-Shot Visual Question Answering 5. Conclusion and References Abstract and 1 Introduction Abstract and 1 Introduction 2. Related Works 2. Related Works PG-Video-LLaVA PG-Video-LLaVA PG-Video-LLaVA 3.1. Overview 3.1. Overview 3.2. Architecture 3.2. Architecture Experiments Experiments Experiments 4.1. Implementation Details 4.1. Implementation Details 4.2. Stronger Baseline 4.2. Stronger Baseline 4.3. Spatial Grounding in Videos 4.3. Spatial Grounding in Videos 4.4. Zero-Shot Visual Question Answering 4.4. Zero-Shot Visual Question Answering 5. Conclusion and References 5. Conclusion and References Supplementary Material Supplementary Material A. Audio Modality Integration B. Visual Grounding: Quantitative Evaluation C. Qualitative Results for Visual Grounding D. Quantitative Evaluations of Video-based Conversation Performance A. Audio Modality Integration A. Audio Modality Integration B. Visual Grounding: Quantitative Evaluation B. Visual Grounding: Quantitative Evaluation C. Qualitative Results for Visual Grounding C. Qualitative Results for Visual Grounding D. Quantitative Evaluations of Video-based Conversation Performance D. Quantitative Evaluations of Video-based Conversation Performance 4.2. Stronger Baseline This section provides an overview of the quantitative evaluations conducted to determine the effects of the strengthened baseline on PG-Video-LLaVA. To evaluate the impact of the enhanced baseline on PG-Video-LLaVA, we apply the benchmarking framework from Video-ChatGPT[22]. This framework measures performance on several axes critical for video-based conversational agents, including correctness of information, detail orientation, contextual understanding, temporal understanding, and consistency. In order to facilitate a reliable and reproducible evaluation, we have updated our assessment pipeline by replacing GPT-3.5-Turbo with Vicuna-13b-v1.5. This adjustment addresses the limitations in reproducibility inherent to the closed-source nature of GPT-3.5-Turbo. Subsequently, we have re-assessed both PG-Video-LLaVA and other recent models to ensure a fair and consistent comparison. The results shown in Table 1 demonstrate that PG-Video-LLaVA outperforms the foundational Video-ChatGPT model and exhibits superior performance when compared to other recent contributions in the domain. Following the quantitative assessment, the qualitative results in Figure 3 indicate the enhanced baseline’s impact on PG-Video-LLaVA’s performance. The PG-Video-LLaVA (13B) model exhibits improved accuracy in the information presented, a deeper level of descriptive detail, and a stronger alignment with the context and temporal progression of the videos. This advancement is particularly noticeable in the precise depiction of the child’s engagement with their surroundings and the giraffe’s behaviour, indicating a refined interpretation of both the activities and their settings. These qualitative insights are consistent with the quantitative results, highlighting the augmented baseline’s role in advancing PG-Video-LLaVA’s capacity in video understanding. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

AI Just Got Better at Watching Videos

Authors:
(1) Albert Gu, Machine Learning Department, Carnegie Mellon University and with equal contribution;
(2) Tri Dao, Department of Computer Science, Princeton University and with equal contribution. Table of Links Abstract and 1 Introduction 2 State Space Models 3 Selective State Space Models and 3.1 Motivation: Selection as a Means of Compression 3.2 Improving SSMs with Selection 3.3 Efficient Implementation of Selective SSMs 3.4 A Simplified SSM Architecture 3.5 Properties of Selection Mechanisms 3.6 Additional Model Details 4 Empirical Evaluation and 4.1 Synthetic Tasks 4.2 Language Modeling 4.3 DNA Modeling 4.4 Audio Modeling and Generation 4.5 Speed and Memory Benchmarks 4.6 Model Ablations 5 Discussion 6 Conclusion and References A Discussion: Selection Mechanism B Related Work C Mechanics of Selective SSMs D Hardware-aware Algorithm For Selective SSMs E Experimental Details and Additional Results 3 Selective State Space Models We motivate our selection mechanism using intuition from synthetic tasks (Section 3.1), then explain how to incorporate this mechanism into state space models (Section 3.2). The resulting time-varying SSMs cannot use convolutions, presenting a technical challenge of how to compute them efficiently. We overcome this with a hardware-aware algorithm that exploits the memory hierarchy on modern hardware (Section 3.3). We then describe a simple SSM architecture without attention or even MLP blocks (Section 3.4). Finally, we discuss some additional properties of selection mechanisms (Section 3.5). 3.1 Motivation: Selection as a Means of Compression We argue that a fundamental problem of sequence modeling is compressing context into a smaller state. In fact, we can view the tradeoffs of popular sequence models from this point of view. For example, attention is both effective and inefficient because it explicitly does not compress context at all. This can be seen from the fact that autoregressive inference requires explicitly storing the entire context (i.e. the KV cache), which directly causes the slow linear-time inference and quadratic-time training of Transformers. On the other hand, recurrent models are efficient because they have a finite state, implying constant-time inference and linear-time training. However, their effectiveness is limited by how well this state has compressed the context. To understand this principle, we focus on two running examples of synthetic tasks (Figure 2). • The Selective Copying task modifies the popular Copying task (Arjovsky, Shah, and Bengio 2016) by varying the position of the tokens to memorize. It requires content-aware reasoning to be able to memorize the relevant tokens (colored) and filter out the irrelevant ones (white). • The Induction Heads task is a well-known mechanism hypothesized to explain the majority of in-context learning abilities of LLMs (Olsson et al. 2022). It requires context-aware reasoning to know when to produce the correct output in the appropriate context (black). In summary, the efficiency vs. effectiveness tradeoff of sequence models is characterized by how well they compress their state: efficient models must have a small state, while effective models must have a state that contains all necessary information from the context. In turn, we propose that a fundamental principle for building sequence models is selectivity: or the context-aware ability to focus on or filter out inputs into a sequential state. In particular, a selection mechanism controls how information propagates or interacts along the sequence dimension (see Section 3.5 for more discussion). This paper is available on arxiv under CC BY 4.0 DEED license. Authors: (1) Albert Gu, Machine Learning Department, Carnegie Mellon University and with equal contribution; (2) Tri Dao, Department of Computer Science, Princeton University and with equal contribution. Authors: Authors: (1) Albert Gu, Machine Learning Department, Carnegie Mellon University and with equal contribution; (2) Tri Dao, Department of Computer Science, Princeton University and with equal contribution. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 2 State Space Models 2 State Space Models 3 Selective State Space Models and 3.1 Motivation: Selection as a Means of Compression 3 Selective State Space Models and 3.1 Motivation: Selection as a Means of Compression 3.2 Improving SSMs with Selection 3.2 Improving SSMs with Selection 3.3 Efficient Implementation of Selective SSMs 3.3 Efficient Implementation of Selective SSMs 3.4 A Simplified SSM Architecture 3.4 A Simplified SSM Architecture 3.5 Properties of Selection Mechanisms 3.5 Properties of Selection Mechanisms 3.6 Additional Model Details 3.6 Additional Model Details 4 Empirical Evaluation and 4.1 Synthetic Tasks 4 Empirical Evaluation and 4.1 Synthetic Tasks 4.2 Language Modeling 4.2 Language Modeling 4.3 DNA Modeling 4.3 DNA Modeling 4.4 Audio Modeling and Generation 4.4 Audio Modeling and Generation 4.5 Speed and Memory Benchmarks 4.5 Speed and Memory Benchmarks 4.6 Model Ablations 4.6 Model Ablations 5 Discussion 5 Discussion 6 Conclusion and References 6 Conclusion and References A Discussion: Selection Mechanism A Discussion: Selection Mechanism B Related Work B Related Work C Mechanics of Selective SSMs C Mechanics of Selective SSMs D Hardware-aware Algorithm For Selective SSMs D Hardware-aware Algorithm For Selective SSMs E Experimental Details and Additional Results E Experimental Details and Additional Results 3 Selective State Space Models We motivate our selection mechanism using intuition from synthetic tasks (Section 3.1), then explain how to incorporate this mechanism into state space models (Section 3.2). The resulting time-varying SSMs cannot use convolutions, presenting a technical challenge of how to compute them efficiently. We overcome this with a hardware-aware algorithm that exploits the memory hierarchy on modern hardware (Section 3.3). We then describe a simple SSM architecture without attention or even MLP blocks (Section 3.4). Finally, we discuss some additional properties of selection mechanisms (Section 3.5). 3.1 Motivation: Selection as a Means of Compression We argue that a fundamental problem of sequence modeling is compressing context into a smaller state. In fact, we can view the tradeoffs of popular sequence models from this point of view. For example, attention is both effective and inefficient because it explicitly does not compress context at all. This can be seen from the fact that autoregressive inference requires explicitly storing the entire context (i.e. the KV cache), which directly causes the slow linear-time inference and quadratic-time training of Transformers. On the other hand, recurrent models are efficient because they have a finite state, implying constant-time inference and linear-time training. However, their effectiveness is limited by how well this state has compressed the context. To understand this principle, we focus on two running examples of synthetic tasks (Figure 2). • The Selective Copying task modifies the popular Copying task (Arjovsky, Shah, and Bengio 2016) by varying the position of the tokens to memorize. It requires content-aware reasoning to be able to memorize the relevant tokens (colored) and filter out the irrelevant ones (white). • The Induction Heads task is a well-known mechanism hypothesized to explain the majority of in-context learning abilities of LLMs (Olsson et al. 2022). It requires context-aware reasoning to know when to produce the correct output in the appropriate context (black). In summary, the efficiency vs. effectiveness tradeoff of sequence models is characterized by how well they compress their state: efficient models must have a small state, while effective models must have a state that contains all necessary information from the context. In turn, we propose that a fundamental principle for building sequence models is selectivity: or the context-aware ability to focus on or filter out inputs into a sequential state. In particular, a selection mechanism controls how information propagates or interacts along the sequence dimension (see Section 3.5 for more discussion). This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

Why Compressing Information Helps AI Work Better

Table of Links Abstract and 1. Introduction Background and Related Work


Threat Model


Robust Style Mimicry


Experimental Setup


Results
6.1 Main Findings: All Protections are Easily Circumvented
6.2 Analysis


Discussion and Broader Impact, Acknowledgements, and References A. Detailed Art Examples B. Robust Mimicry Generations C. Detailed Results D. Differences with Glaze Finetuning E. Findings on Glaze 2.0 F. Findings on Mist v2 G. Methods for Style Mimicry H. Existing Style Mimicry Protections I. Robust Mimicry Methods J. Experimental Setup K. User Study L. Compute Resources D Differences with Glaze Finetuning In Section 4.1 and Figure 2, we discussed the brittleness of Glaze protections against small changes in the finetuning script. We also found our finetuning setup to be better at baseline style mimicry from unprotected art (see Figure 19). Authors:
(1) Robert Honig, ETH Zurich (robert.hoenig@inf.ethz.ch);
(2) Javier Rando, ETH Zurich (javier.rando@inf.ethz.ch);
(3) Nicholas Carlini, Google DeepMind;
(4) Florian Tramer, ETH Zurich (florian.tramer@inf.ethz.ch). This paper is available on arxiv under CC BY 4.0 license. Table of Links Abstract and 1. Introduction Abstract and 1. Introduction Background and Related Work Threat Model Robust Style Mimicry Experimental Setup Results
6.1 Main Findings: All Protections are Easily Circumvented
6.2 Analysis Discussion and Broader Impact, Acknowledgements, and References Background and Related Work Background and Related Work Background and Related Work Threat Model Threat Model Threat Model Robust Style Mimicry Robust Style Mimicry Robust Style Mimicry Experimental Setup Experimental Setup Experimental Setup Results 6.1 Main Findings: All Protections are Easily Circumvented 6.2 Analysis Results Results 6.1 Main Findings: All Protections are Easily Circumvented 6.1 Main Findings: All Protections are Easily Circumvented 6.2 Analysis 6.2 Analysis Discussion and Broader Impact, Acknowledgements, and References Discussion and Broader Impact, Acknowledgements, and References Discussion and Broader Impact, Acknowledgements, and References A. Detailed Art Examples A. Detailed Art Examples B. Robust Mimicry Generations B. Robust Mimicry Generations C. Detailed Results C. Detailed Results D. Differences with Glaze Finetuning D. Differences with Glaze Finetuning E. Findings on Glaze 2.0 E. Findings on Glaze 2.0 F. Findings on Mist v2 F. Findings on Mist v2 G. Methods for Style Mimicry G. Methods for Style Mimicry H. Existing Style Mimicry Protections H. Existing Style Mimicry Protections I. Robust Mimicry Methods I. Robust Mimicry Methods J. Experimental Setup J. Experimental Setup K. User Study K. User Study L. Compute Resources L. Compute Resources D Differences with Glaze Finetuning In Section 4.1 and Figure 2, we discussed the brittleness of Glaze protections against small changes in the finetuning script. We also found our finetuning setup to be better at baseline style mimicry from unprotected art (see Figure 19). Authors: (1) Robert Honig, ETH Zurich (robert.hoenig@inf.ethz.ch); (2) Javier Rando, ETH Zurich (javier.rando@inf.ethz.ch); (3) Nicholas Carlini, Google DeepMind; (4) Florian Tramer, ETH Zurich (florian.tramer@inf.ethz.ch). Authors: Authors: (1) Robert Honig, ETH Zurich (robert.hoenig@inf.ethz.ch); (2) Javier Rando, ETH Zurich (javier.rando@inf.ethz.ch); (3) Nicholas Carlini, Google DeepMind; (4) Florian Tramer, ETH Zurich (florian.tramer@inf.ethz.ch). This paper is available on arxiv under CC BY 4.0 license. This paper is available on arxiv under CC BY 4.0 license. available on arxiv available on arxiv

The Key to Better Art Style Mimicry

Hey Hackers, Hidonix has been nominated in HackerNoon's annual Startups of The Year awards in Los Angeles, California. Please vote for us here: https://hackernoon.com/startups/north-america/north-america-los-angeles-ca-usa?stup=66f3161b6ea4677e23d79637&embedable=true Read more about us below to understand why we deserve your vote. Meet Hidonix At first glance, you might think we’re just another software company. But we’re not. We’re a deep tech company dedicated to investing in R&D across groundbreaking technologies: Extended Reality, Computer Vision, Artificial Intelligence, Spatial Intelligence, Robotics, and IoT. Our mission is to push the boundaries of these technologies and integrate them into innovative, in-house platforms that we provide to various industries, including museums, congress venues, malls, hospitals, and schools. The best way to visualize who we are is through our lab—a space where we experiment with these technologies to deliver solutions that have never been achieved before. https://youtu.be/FKIl_0ua2j0?embedable=true How Hidonix achieved product-market fit One of the most innovative aspects of Hidonix, and something both our customers and investors deeply value, is the versatility of our technologies. While they are adaptable across many verticals, they are also tailored to meet the specific needs of the industries we target. We achieve this by embedding our technologies into purpose-built products designed specifically for each market. For example, our Computer Vision algorithms enable museums to deliver detailed information about their collections through a simple scan of the artworks. Meanwhile, in the education sector, we use the same cutting-edge algorithms to perform advanced facial recognition, helping schools prevent potential threats and ensure safety. The evolution of Hidonix Hidonix began in the museum field, where we successfully introduced innovative, interactive solutions to attract and engage new visitors. At the same time, we empowered internal staff with an integrated platform offering over 60 services to streamline their daily operations. These include managing artwork loans and preservation, digitizing collections, and building a comprehensive CRM. Initially, our technologies were designed specifically for museums. However, we soon realized that the same technologies could be adapted to other industries. Our indoor navigation system, for example, proved highly versatile. This discovery led us to explore and address the unique needs of various verticals by tailoring our algorithms to fit specific business requirements. This process transformed us from a small company in Italy into a financially stable organization with a team of 30+ professionals spread across Europe and the U.S., where our headquarters are now located. Milestones and achievements We’re proud of the many projects that demonstrate our expertise and innovation. Among the highlights are our collaborations with the Venice Biennale, where we supported the Venice Pavilion during the challenging COVID-19 period. Our solutions helped welcome, educate, and entertain visitors at a time when cultural initiatives struggled to engage audiences without physical attendance. Another notable achievement is our work with one of the largest convention centers in Europe. By implementing our navigation feature, we not only enhanced the visitor experience but also created new revenue streams for exhibitors and event organizers. Of course, there’s still much more to achieve. We’re currently focused on exciting U.S. projects that address critical issues like safety and security in schools. At Hidonix, we’re committed to continuous R&D, confident in our ability to make a meaningful impact across diverse sectors. Lessons learned When working with advanced technologies, we’ve realized that the market isn’t always immediately ready to embrace them. It takes time for people to understand how to best use new solutions and become familiar with innovative concepts. Not all of us are Steve Jobs—convincing others to believe in your vision can be challenging at first. To all the innovators out there: stay consistent and don’t be discouraged by initial feedback. Keep pushing forward, and if you’re truly making a real impact, people will recognize it eventually. What Startups of The Year means to us We are incredibly proud to be nominated by the community—it has strengthened our belief in what we’re building. As a team of young and talented individuals, we’re often amazed by what we can achieve when we think outside the box, driven by the hope that our technologies will make a meaningful impact on society, not just generate profits. Winning this competition would be an incredible honor and a well-deserved recognition for every team member who has put in unparalleled effort into our mission. It would also help us amplify our message about innovative ways to use technology in fields that urgently need advancement for the greater good. Conclusion Being part of the "Startups of the Year" journey is a humbling and inspiring experience for us. It reinforces our commitment to developing innovative technologies that create real value and impact across industries. This recognition, whether we win or not, fuels our passion to keep pushing boundaries, thinking creatively, and striving for excellence. We are deeply grateful for the support of our community, our talented team, and everyone who believes in our mission. Together, we’re not just building a company—we’re shaping a future where technology truly benefits society! Vote for us today! Welcome to HackerNoon’s Startups of The Year 2024 interview series, spotlighting startups from around the world that are changing the way we live, one solution at a time. If you’ve been nominated, create a brand profile and answer these questions here.
Want to nominate a startup? Follow the steps here.Looking for inspiration? Check out how other startups responded to last year’s Startups of The Year interview questions and examples of company responses here, here, and here. Wanna use this template? Click here. Hey Hackers, Hidonix has been nominated in HackerNoon's annual Startups of The Year awards in Los Angeles, California. Please vote for us here: Please vote for us here: https://hackernoon.com/startups/north-america/north-america-los-angeles-ca-usa?stup=66f3161b6ea4677e23d79637&embedable=true https://hackernoon.com/startups/north-america/north-america-los-angeles-ca-usa?stup=66f3161b6ea4677e23d79637&embedable=true Read more about us below to understand why we deserve your vote. Meet Hidonix At first glance, you might think we’re just another software company. But we’re not. We’re a deep tech company dedicated to investing in R&D across groundbreaking technologies: Extended Reality, Computer Vision, Artificial Intelligence, Spatial Intelligence, Robotics, and IoT. Our mission is to push the boundaries of these technologies and integrate them into innovative, in-house platforms that we provide to various industries, including museums, congress venues, malls, hospitals, and schools. The best way to visualize who we are is through our lab—a space where we experiment with these technologies to deliver solutions that have never been achieved before. https://youtu.be/FKIl_0ua2j0?embedable=true https://youtu.be/FKIl_0ua2j0?embedable=true How Hidonix achieved product-market fit One of the most innovative aspects of Hidonix , and something both our customers and investors deeply value, is the versatility of our technologies. While they are adaptable across many verticals, they are also tailored to meet the specific needs of the industries we target. We achieve this by embedding our technologies into purpose-built products designed specifically for each market. Hidonix For example, our Computer Vision algorithms enable museums to deliver detailed information about their collections through a simple scan of the artworks. Meanwhile, in the education sector, we use the same cutting-edge algorithms to perform advanced facial recognition, helping schools prevent potential threats and ensure safety. The evolution of Hidonix Hidonix began in the museum field, where we successfully introduced innovative, interactive solutions to attract and engage new visitors. At the same time, we empowered internal staff with an integrated platform offering over 60 services to streamline their daily operations. These include managing artwork loans and preservation, digitizing collections, and building a comprehensive CRM. Initially, our technologies were designed specifically for museums. However, we soon realized that the same technologies could be adapted to other industries. Our indoor navigation system, for example, proved highly versatile. This discovery led us to explore and address the unique needs of various verticals by tailoring our algorithms to fit specific business requirements. This process transformed us from a small company in Italy into a financially stable organization with a team of 30+ professionals spread across Europe and the U.S., where our headquarters are now located. Milestones and achievements We’re proud of the many projects that demonstrate our expertise and innovation. Among the highlights are our collaborations with the Venice Biennale , where we supported the Venice Pavilion during the challenging COVID-19 period. Our solutions helped welcome, educate, and entertain visitors at a time when cultural initiatives struggled to engage audiences without physical attendance. Venice Biennale Another notable achievement is our work with one of the largest convention centers in Europe. By implementing our navigation feature, we not only enhanced the visitor experience but also created new revenue streams for exhibitors and event organizers. Of course, there’s still much more to achieve. We’re currently focused on exciting U.S. projects that address critical issues like safety and security in schools. At Hidonix, we’re committed to continuous R&D, confident in our ability to make a meaningful impact across diverse sectors. Lessons learned When working with advanced technologies, we’ve realized that the market isn’t always immediately ready to embrace them. It takes time for people to understand how to best use new solutions and become familiar with innovative concepts. Not all of us are Steve Jobs—convincing others to believe in your vision can be challenging at first. To all the innovators out there: stay consistent and don’t be discouraged by initial feedback. Keep pushing forward, and if you’re truly making a real impact, people will recognize it eventually. What Startups of The Year means to us What Startups of The Year means to us We are incredibly proud to be nominated by the community—it has strengthened our belief in what we’re building. As a team of young and talented individuals, we’re often amazed by what we can achieve when we think outside the box, driven by the hope that our technologies will make a meaningful impact on society, not just generate profits. Winning this competition would be an incredible honor and a well-deserved recognition for every team member who has put in unparalleled effort into our mission. It would also help us amplify our message about innovative ways to use technology in fields that urgently need advancement for the greater good. Conclusion Being part of the "Startups of the Year" journey is a humbling and inspiring experience for us. It reinforces our commitment to developing innovative technologies that create real value and impact across industries. This recognition, whether we win or not, fuels our passion to keep pushing boundaries, thinking creatively, and striving for excellence. We are deeply grateful for the support of our community, our talented team, and everyone who believes in our mission. Together, we’re not just building a company—we’re shaping a future where technology truly benefits society! Vote for us today! Vote for us today! Vote for us today! Welcome to HackerNoon’s Startups of The Year 2024 interview series, spotlighting startups from around the world that are changing the way we live, one solution at a time. If you’ve been nominated, create a brand profile and answer these questions here. Want to nominate a startup? Follow the steps here. Looking for inspiration? Check out how other startups responded to last year’s Startups of The Year interview questions and examples of company responses here, here, and here. Welcome to HackerNoon’s Startups of The Year 2024 interview series, spotlighting startups from around the world that are changing the way we live, one solution at a time. If you’ve been nominated, create a brand profile and answer these questions here . Startups of The Year 2024 create a brand profile create a brand profile create a brand profile here here Want to nominate a startup? Follow the steps here . here here Looking for inspiration? Check out how other startups responded to last year’s Startups of The Year interview questions and examples of company responses here, here, and here. Looking for inspiration? Check out how other startups responded to last year’s Startups of The Year interview questions and examples of company responses here , here , and here . last year’s Startups of The Year interview questions last year’s Startups of The Year interview questions here here here here here here Wanna use this template? Click here. Wanna use this template? Click here . here

Meet the Startup Changing Museums For the Better

ለተሻለ የጅምር ለውጥ ሙዚየሞችን ያግኙ

Upoznajte startup koji mijenja muzeje na bolje

Lernen Sie das Startup kennen, das Museen zum Besseren verändert

Conozca la startup que está cambiando los museos para mejor

博物館をより良い方向に変えるスタートアップ企業

박물관을 더 나은 방향으로 바꾸는 스타트업을 만나보세요

Susipažinkite su startuoliu, keičiančiu muziejus į gerąją pusę

Hihaona amin'ireo tranombakoka fanovana fanombohana ho amin'ny tsaratsara kokoa

Maak kennis met de startup die musea ten goede verandert

Kilalanin ang Startup Changing Museums For the Better

Зустрічайте стартап, який змінює музеї на краще

Muzeylarni yaxshi tomonga o'zgartiruvchi startap bilan tanishing

Marvel Rivals, released on December 6, 2024, on Xbox, PC, and PlayStation, is the latest entry in the hero shooter genre. While Overwatch has long dominated this space, it may be time for a new chaotic, team-based game to take the spotlight. East and West Synergy Developed by gaming giant NetEase Games, headquartered in Guangzhou, China, with offices in Seattle, the game utilizes iconic Marvel superheroes like Spider-Man, Wolverine, and Captain America, while also allowing players to team up with villains like Venom, Magneto, and Loki in action-packed, colorful 6v6 arenas. The game boasts excellent sound design, well-crafted maps, and simple gameplay that players familiar with the genre will quickly pick up. Being a third-person shooter, Marvel Rivals showcases each Marvel legend in their iconic glory. Each character stays true to their source material, offering expressive, engaging designs that are fun to watch and play. What stands out most to me, beyond the captivating gameplay and vibrant art style, is the potential for Western and Eastern audiences to come together through characters beloved for over 60 years. Spider-Man, the Hulk — these are universally recognized figures. Even though the characters were created in America, the Marvel Universe enjoys a massive following in China, highlighted by the Marvel Universe attraction at Shanghai Disneyland. China is the largest gaming market in the world, making the financial incentive for a game like Marvel Rivals undeniable. With dedicated servers globally, the game has strong potential to unite Eastern and Western communities while fostering competition through esports. Although we don’t yet know the full roadmap for competitive esports within the game, the Marvel trademark could prove invaluable in securing big prize pools that entice players to compete for trophies and profit. https://youtu.be/8NUtV1CkyKs?embedable=true The inclusion of Tokyo-themed stages in the game is a nice touch, drawing attention to the world’s love of Japanese culture and appealing to Marvel’s Japanese fanbase. If NetEase manages this project well, Marvel Rivals could become one of the biggest games, connecting Western and Eastern fanbases, casual players, and aspiring professionals alike. Perhaps we’ll even see Chinese players gain recognition in the West for their incredible gameplay as Rocket Raccoon or Iron Man. The game even features some Chinese writing, which has left some players confused. This inclusion of Chinese characters is likely part of a broader strategy to celebrate and bring attention to Chinese culture and artists. Privacy Concerns and Political Implications If you’ve been following the news, you might know that TikTok faces a potential ban at the beginning of 2025 due to “national security” risks. I wonder if Marvel Rivals and NetEase will face similar scrutiny regarding privacy concerns, given that the company is based in China. Such accusations could have significant implications for the future of gaming. It’s already well-known that games today collect vast amounts of data on players. Perhaps this concern is premature, but it’s worth considering how technologies influence public relations and politics. Still, I believe accusations of serious privacy violations or concerns about how a Chinese company uses player data will likely go ignored. Most players care little about the political and economic implications of games as long as their gameplay isn’t affected or the costs of in-game cosmetics and DLC don’t rise. Of course, there have been instances where politics, activism, and esports have collided. The more connected the world becomes, the more issues are spotlighted and shared. Questions for the Future How will NetEase’s origins impact its ability to secure sponsorships? How much will China’s gaming industry invest in making Marvel Rivals a viable product in the competitive gaming space? Will international teams be brought in? Could major players in esports, like Saudi Arabia, get involved? There’s potential for Marvel Rivals to feature at the World Esports Cup in 2025, but many questions remain. Nevertheless, this game could serve as a valuable tool to unite global gaming culture and add fuel to the rapidly growing esports machine. China has actively promoted the global dissemination of its games as a cultural export, with several games enjoying major success abroad. The country has established a policy framework to promote traditional Chinese culture through innovative digital mediums, including high-quality games. In 2019, Beijing unveiled gaming industry development guidelines that aim to leverage games as an international medium for conveying compelling Chinese stories.
— Liz Wang, The Role of Gaming and Esports in China’s Innovation Strategy: A Cultural Export and Source of Soft Power, August 2024 Marvel Rivals, released on December 6, 2024, on Xbox, PC, and PlayStation, is the latest entry in the hero shooter genre. While Overwatch has long dominated this space, it may be time for a new chaotic, team-based game to take the spotlight. Marvel Rivals, Overwatch East and West Synergy Developed by gaming giant NetEase Games, headquartered in Guangzhou, China, with offices in Seattle, the game utilizes iconic Marvel superheroes like Spider-Man, Wolverine, and Captain America, while also allowing players to team up with villains like Venom, Magneto, and Loki in action-packed, colorful 6v6 arenas. The game boasts excellent sound design, well-crafted maps, and simple gameplay that players familiar with the genre will quickly pick up. Being a third-person shooter, Marvel Rivals showcases each Marvel legend in their iconic glory. Each character stays true to their source material, offering expressive, engaging designs that are fun to watch and play. Marvel Rivals What stands out most to me, beyond the captivating gameplay and vibrant art style, is the potential for Western and Eastern audiences to come together through characters beloved for over 60 years. Spider-Man, the Hulk — these are universally recognized figures. Even though the characters were created in America, the Marvel Universe enjoys a massive following in China, highlighted by the Marvel Universe attraction at Shanghai Disneyland. Marvel Universe China is the largest gaming market in the world, making the financial incentive for a game like Marvel Rivals undeniable. With dedicated servers globally, the game has strong potential to unite Eastern and Western communities while fostering competition through esports. Marvel Rivals Although we don’t yet know the full roadmap for competitive esports within the game, the Marvel trademark could prove invaluable in securing big prize pools that entice players to compete for trophies and profit. https://youtu.be/8NUtV1CkyKs?embedable=true https://youtu.be/8NUtV1CkyKs?embedable=true The inclusion of Tokyo-themed stages in the game is a nice touch, drawing attention to the world’s love of Japanese culture and appealing to Marvel’s Japanese fanbase. If NetEase manages this project well, Marvel Rivals could become one of the biggest games, connecting Western and Eastern fanbases, casual players, and aspiring professionals alike. Marvel Rivals Perhaps we’ll even see Chinese players gain recognition in the West for their incredible gameplay as Rocket Raccoon or Iron Man. The game even features some Chinese writing, which has left some players confused. This inclusion of Chinese characters is likely part of a broader strategy to celebrate and bring attention to Chinese culture and artists. Privacy Concerns and Political Implications Privacy Concerns and Political Implications If you’ve been following the news, you might know that TikTok faces a potential ban at the beginning of 2025 due to “national security” risks. I wonder if Marvel Rivals and NetEase will face similar scrutiny regarding privacy concerns, given that the company is based in China. Such accusations could have significant implications for the future of gaming. Marvel Rivals It’s already well-known that games today collect vast amounts of data on players. Perhaps this concern is premature, but it’s worth considering how technologies influence public relations and politics. Still, I believe accusations of serious privacy violations or concerns about how a Chinese company uses player data will likely go ignored. Most players care little about the political and economic implications of games as long as their gameplay isn’t affected or the costs of in-game cosmetics and DLC don’t rise. Of course, there have been instances where politics, activism, and esports have collided. The more connected the world becomes, the more issues are spotlighted and shared. Questions for the Future Questions for the Future How will NetEase’s origins impact its ability to secure sponsorships? How much will China’s gaming industry invest in making Marvel Rivals a viable product in the competitive gaming space? Will international teams be brought in? Could major players in esports, like Saudi Arabia, get involved? Marvel Rivals There’s potential for Marvel Rivals to feature at the World Esports Cup in 2025, but many questions remain. Nevertheless, this game could serve as a valuable tool to unite global gaming culture and add fuel to the rapidly growing esports machine. Marvel Rivals China has actively promoted the global dissemination of its games as a cultural export, with several games enjoying major success abroad. The country has established a policy framework to promote traditional Chinese culture through innovative digital mediums, including high-quality games. In 2019, Beijing unveiled gaming industry development guidelines that aim to leverage games as an international medium for conveying compelling Chinese stories. — Liz Wang, The Role of Gaming and Esports in China’s Innovation Strategy: A Cultural Export and Source of Soft Power, August 2024 China has actively promoted the global dissemination of its games as a cultural export, with several games enjoying major success abroad. The country has established a policy framework to promote traditional Chinese culture through innovative digital mediums, including high-quality games. In 2019, Beijing unveiled gaming industry development guidelines that aim to leverage games as an international medium for conveying compelling Chinese stories. China has actively promoted the global dissemination of its games as a cultural export, with several games enjoying major success abroad. The country has established a policy framework to promote traditional Chinese culture through innovative digital mediums, including high-quality games. In 2019, Beijing unveiled gaming industry development guidelines that aim to leverage games as an international medium for conveying compelling Chinese stories. — Liz Wang, The Role of Gaming and Esports in China’s Innovation Strategy: A Cultural Export and Source of Soft Power, August 2024 — Liz Wang, The Role of Gaming and Esports in China’s Innovation Strategy: A Cultural Export and Source of Soft Power, August 2024

Could Marvel Rivals Help Facilitate Better US-China Relations?

The fever around AI has many founders asking whether they should let an LLM like ChatGPT take a crack at their pitch deck. After all, if AI can write code and create marketing copy, why not trust it with your startup's story? In Short: No LLMs are essentially sophisticated pattern matchers, but I would argue unsuitable for breakthrough ideas. These models work by predicting what words should come next based on the statistical patterns in their training data – they're essentially playing a complex game of "finish this sentence." And while they might seem magical, they're ultimately bound by the patterns they've seen before, creating legible remixes rather than truly novel thought. Most models try to fake insight with a dash of randomness in their output (what AI researchers call "temperature"), but even this randomness is still just pulling from their statistical set. Your startup’s idea is the opposite of pattern matching. You're building something new, something based on your real-world insight that will break patterns. Something that will disrupt. Think about how a founder identifies a real opportunity: they notice friction in daily life, spot inefficiencies in their industry, or experience a problem firsthand that no one else has solved. This kind of pattern breaking – seeing what's missing rather than what's already there – is precisely what LLMs can't do. They can tell you all about successful companies that exist, but they can't identify the gaps that will lead to the next breakthrough. Regardless of the effectiveness of AI, creating a pitch deck isn't just about the end product. It's a mental exercise for founders to crystallize their thinking about their market, users, and competitive landscape. Skipping this process by outsourcing it to AI is like trying to get in shape by having someone else do your pushups. The real value isn't in the deck itself – it's in the mental model you build while creating it. After working with countless founders on their decks, time and time again I hear that the shiny slides were just half the value, and an often better takeaway from fundraising is the deeper understanding of their venture and how it fits into a changing world. Even if AI gets good enough in the next few years to truly create great decks, I'd still recommend founders go manual so they can experience the deep dive. Where AI Actually Makes Sense That said, AI isn't entirely useless in the pitch process. Consider using it as an occasional tool, not a crutch. Some founders have the opposite problem of having too little to say – they have too much. When you're deep in the trenches of building, it's easy to write paragraphs where sentences or even half sentences would do. This is where AI shines: as an efficient editor. Feed it any verbose slide, and it can help distill your message to its essential elements while preserving your core insights. Founder bios are an obvious candidate for this treatment – AI can help transform a meandering career history into a sharp narrative about why they’re the right person to solve this specific problem. The same approach works for market analysis slides that read like research papers, or product descriptions that get lost in technical details. Again, the mental exercise of cutting down text yourself is probably more useful in helping you understand what really matters for your business. But using AI in this way at least avoids the issues of hallucination or lacking your novel insights. Once you feel your deck is ready for primetime, AI can be used to review and practice your pitch. Tools like Fornax.ai can analyze your deck against common pitch deck tactics, highlighting where you might be breaking conventional wisdom (in good or bad ways). Additionally, platforms like Yoodli can help you with verbal pitch practice by providing AI feedback on everything from your pace to your handling of questions. The key is remembering that AI can help you refine your pitch, not create it. Your unique insights, experience, and deep understanding of the problem you're solving – that's still on you. The fever around AI has many founders asking whether they should let an LLM like ChatGPT take a crack at their pitch deck. After all, if AI can write code and create marketing copy, why not trust it with your startup's story? In Short: No LLMs are essentially sophisticated pattern matchers, but I would argue unsuitable for breakthrough ideas. These models work by predicting what words should come next based on the statistical patterns in their training data – they're essentially playing a complex game of "finish this sentence." And while they might seem magical, they're ultimately bound by the patterns they've seen before, creating legible remixes rather than truly novel thought. Most models try to fake insight with a dash of randomness in their output (what AI researchers call "temperature"), but even this randomness is still just pulling from their statistical set. Your startup’s idea is the opposite of pattern matching. You're building something new, something based on your real-world insight that will break patterns. Something that will disrupt. Think about how a founder identifies a real opportunity: they notice friction in daily life, spot inefficiencies in their industry, or experience a problem firsthand that no one else has solved. This kind of pattern breaking – seeing what's missing rather than what's already there – is precisely what LLMs can't do. They can tell you all about successful companies that exist, but they can't identify the gaps that will lead to the next breakthrough. Regardless of the effectiveness of AI, creating a pitch deck isn't just about the end product. It's a mental exercise for founders to crystallize their thinking about their market, users, and competitive landscape. Skipping this process by outsourcing it to AI is like trying to get in shape by having someone else do your pushups. The real value isn't in the deck itself – it's in the mental model you build while creating it. After working with countless founders on their decks , time and time again I hear that the shiny slides were just half the value, and an often better takeaway from fundraising is the deeper understanding of their venture and how it fits into a changing world. Even if AI gets good enough in the next few years to truly create great decks, I'd still recommend founders go manual so they can experience the deep dive. After working with countless founders on their decks Where AI Actually Makes Sense That said, AI isn't entirely useless in the pitch process. Consider using it as an occasional tool, not a crutch. Some founders have the opposite problem of having too little to say – they have too much. When you're deep in the trenches of building, it's easy to write paragraphs where sentences or even half sentences would do. This is where AI shines: as an efficient editor. Feed it any verbose slide, and it can help distill your message to its essential elements while preserving your core insights. Founder bios are an obvious candidate for this treatment – AI can help transform a meandering career history into a sharp narrative about why they’re the right person to solve this specific problem. The same approach works for market analysis slides that read like research papers, or product descriptions that get lost in technical details. Again, the mental exercise of cutting down text yourself is probably more useful in helping you understand what really matters for your business. But using AI in this way at least avoids the issues of hallucination or lacking your novel insights. Once you feel your deck is ready for primetime, AI can be used to review and practice your pitch. Tools like Fornax.ai can analyze your deck against common pitch deck tactics, highlighting where you might be breaking conventional wisdom (in good or bad ways). Additionally, platforms like Yoodli can help you with verbal pitch practice by providing AI feedback on everything from your pace to your handling of questions. The key is remembering that AI can help you refine your pitch, not create it. Your unique insights, experience, and deep understanding of the problem you're solving – that's still on you.

Your Startup Deserves Better Than AI’s Pattern-Matching Pitches

የእርስዎ ጅምር ከ AI ጥለት-ተዛማጅ እርከኖች የተሻለ ይገባዋል

Ihr Startup verdient mehr als die Mustererkennungs-Pitches der KI

Su startup merece algo mejor que los discursos de búsqueda de patrones de la IA

A túa startup merece algo mellor que os lanzamentos de coincidencia de patróns da IA

あなたのスタートアップは、AI のパターンマッチングのピッチよりも優れたものに値する

თქვენი სტარტაპი უკეთესს იმსახურებს, ვიდრე AI-ის შაბლონების შესატყვისი მოედანი

ការចាប់ផ្តើមរបស់អ្នកសមនឹងទទួលបានប្រសើរជាង AI's Pattern-Matching Pitches

귀하의 스타트업은 AI의 패턴 매칭 피치보다 더 나은 것을 받을 자격이 있습니다.

Jūsų startuolis nusipelno geresnių nei dirbtinio intelekto modelių atitikimo pasiūlymai

Váš startup si zaslúži lepšie ako umelá inteligencia na výber vzorov

Оғози шумо сазовортар аз сатҳҳои мувофиқи намунаи AI аст

สตาร์ทอัพของคุณสมควรได้รับสิ่งที่ดีกว่าการนำเสนอการจับคู่รูปแบบของ AI

As a PhD student, you’re likely feeling a mix of excitement and overwhelm as you embark on your research journey. Reflecting on my own experience, I remember the challenges and limitations of the tools available on the market to assist with conducting your literature review. During the COVID period, tools like Research Rabbit and Zotero were a bit more rudimentary and easier to use, and they proved invaluable in my first year. However, this realisation prompted me to share insights on how to effectively utilise such tools. This short blog was inspired after watching the potential of AI tools showcased in “Write A Masterpiece Systematic Literature Review With AI [Next Level Strategies]” by Andy Stapleton, Please note that the advice and approaches discussed in this blog reflect my personal experience and are not professional academic advice. In this blog, I will introduce the Triple T Approach — Tactics, Techniques, and Tools — to streamline literature reviews. By integrating strategic planning, methodological execution, and AI tools like ChatGPT, Elicit, and Litmap, this approach ensures a comprehensive and efficient review process. Be Aware of the Texas sharpshooter! Before diving into the Triple T framework, it’s crucial to heed the advice of being wary of the Texas sharpshooter fallacy. that refers to a scenario where a person, lacking actual shooting skills, randomly fires bullets at a barn wall. Subsequently, they paint a bull’s-eye around a fortuitous cluster of bullet holes and proclaim themselves a sharpshooter. Falling into this trap can be tempting, particularly with tools that produce seemingly strong evidence, which may obscure their biases. This is why it’s essential to question everything produced by AI, by adopting a philosophical skepticism mindset. It’s crucial to question everything produced by AI to provoke awareness of any bias existing in the data set and challenge our assumptions while evaluating evidence generated by GenAI tools and exploring different perspectives and their implications on our research direction. This logical pitfall warns against being drawn to tools that appear to provide strong evidence, potentially concealing biases. Utilizing these tools effectively requires critical thinking skills; in this context, these skills act as your shooting skills for the Texas sharpshooter fallacy! This is why I always recommend reading more about this topic, as it introduces essential life skills to help not only grow your research but also grow as a researcher. This approach ensures a comprehensive and efficient review process while also keeping in mind the limitations of current AI tools and their biases. When choosing each tactic, try to link it with appropriate tools and techniques while considering how they align together and the implications of such an approach on the limitations of AI tools. I have only focused on the bias challenge, but of course, there are many limitations awaiting AI in academia. Always remember that you are leading the tools, not the other way around. What is the Triple T Approach? The Triple T framework breaks down into Tactics, Techniques, and Tools. Tactics involve strategic decisions in structuring the review, conducting thorough literature searches, and synthesizing information. Techniques encompass methods such as AI-assisted structuring and analytical reading, while Tools consist of resources like ChatGPT and Mendeley for efficient literature management. Applying the Triple T Approach to Your Literature Review Workflow: Here’s a quick walkthrough of some of the concepts and strategies applied in the approach to ensure a methodical and tool-driven process, enhancing the quality and depth of your review. 1. Tactics: • Structuring the Review: Develop a clear, organized structure for your literature review. • Comprehensive Literature Search: Systematically search for and select relevant literature. • Synthesizing Information: Integrating findings from various sources into a cohesive narrative. 2. Techniques: • AI-Assisted Structuring: Using AI to generate a preliminary structure for your review. • Targeted Literature Discovery: Employing specific queries based on your review structure to find relevant literature. • Analytical Reading and Organization: Critically reading and categorizing literature for easy reference and synthesis. 3. Tools: • ChatGPT and Perplexity: For generating the initial structure of your literature review. • PaperPal and Jenni AI: Tools to assist with proofreading your initial drafts and refining your literature structure. • Elicit or Scispace: To search for and summarize seed papers based on your structured queries. • Litmap and Connected Papers: For discovering related literature reviews. • Mendeley or Zetro: To manage and organize your references and PDFs. • Docanalyzer.AI: For analyzing documents and aligning them with your review structure. By aligning your research with this approach, you’ll streamline your workflow and lay a robust foundation for your literature review paper. Of course, this approach is flexible. For example, when you start structuring the plan for your literature review, you could use tools like Perplexity and then draft and proofread the initial paper structure using PaperPal. However, you could also gather information before you do any planning. Thus, a tactic like synthesizing information will need some techniques for targeted literature discovery through Elicit. So, please add or update any section of this approach that you feel fits your needs! As a PhD student, you’re likely feeling a mix of excitement and overwhelm as you embark on your research journey. Reflecting on my own experience, I remember the challenges and limitations of the tools available on the market to assist with conducting your literature review. During the COVID period, tools like Research Rabbit and Zotero were a bit more rudimentary and easier to use, and they proved invaluable in my first year. However, this realisation prompted me to share insights on how to effectively utilise such tools. This short blog was inspired after watching the potential of AI tools showcased in “ Write A Masterpiece Systematic Literature Review With AI [Next Level Strategies ]” by Andy Stapleton, Please note that the advice and approaches discussed in this blog reflect my personal experience and are not professional academic advice. In this blog, I will introduce the Triple T Approach — Tactics, Techniques, and Tools — to streamline literature reviews. By integrating strategic planning, methodological execution, and AI tools like ChatGPT, Elicit, and Litmap, this approach ensures a comprehensive and efficient review process. Write A Masterpiece Systematic Literature Review With AI [Next Level Strategies Be Aware of the Texas sharpshooter! Be Aware of the Texas sharpshooter! Before diving into the Triple T framework, it’s crucial to heed the advice of being wary of the Texas sharpshooter fallacy . that refers to a scenario where a person, lacking actual shooting skills, randomly fires bullets at a barn wall. Subsequently, they paint a bull’s-eye around a fortuitous cluster of bullet holes and proclaim themselves a sharpshooter. Falling into this trap can be tempting, particularly with tools that produce seemingly strong evidence, which may obscure their biases. Texas sharpshooter fallacy This is why it’s essential to question everything produced by AI, by adopting a philosophical skepticism mindset. It’s crucial to question everything produced by AI to provoke awareness of any bias existing in the data set and challenge our assumptions while evaluating evidence generated by GenAI tools and exploring different perspectives and their implications on our research direction. This logical pitfall warns against being drawn to tools that appear to provide strong evidence, potentially concealing biases. Utilizing these tools effectively requires critical thinking skills; in this context, these skills act as your shooting skills for the Texas sharpshooter fallacy! This is why I always recommend reading more about this topic, as it introduces essential life skills to help not only grow your research but also grow as a researcher. This approach ensures a comprehensive and efficient review process while also keeping in mind the limitations of current AI tools and their biases. When choosing each tactic, try to link it with appropriate tools and techniques while considering how they align together and the implications of such an approach on the limitations of AI tools. I have only focused on the bias challenge, but of course, there are many limitations awaiting AI in academia. Always remember that you are leading the tools, not the other way around. you are leading What is the Triple T Approach? What is the Triple T Approach? The Triple T framework breaks down into Tactics, Techniques, and Tools. Tactics involve strategic decisions in structuring the review, conducting thorough literature searches, and synthesizing information. Techniques encompass methods such as AI-assisted structuring and analytical reading, while Tools consist of resources like ChatGPT and Mendeley for efficient literature management. strategic decisions methods resources Applying the Triple T Approach to Your Literature Review Workflow: Applying the Triple T Approach to Your Literature Review Workflow: Here’s a quick walkthrough of some of the concepts and strategies applied in the approach to ensure a methodical and tool-driven process, enhancing the quality and depth of your review. 1. Tactics: 1. Tactics: • Structuring the Review : Develop a clear, organized structure for your literature review. Structuring the Review • Comprehensive Literature Search : Systematically search for and select relevant literature. Comprehensive Literature Search • Synthesizing Information : Integrating findings from various sources into a cohesive narrative. Synthesizing Information 2. Techniques: 2. Techniques: • AI-Assisted Structuring : Using AI to generate a preliminary structure for your review. AI-Assisted Structuring • Targeted Literature Discovery : Employing specific queries based on your review structure to find relevant literature. Targeted Literature Discovery • Analytical Reading and Organization : Critically reading and categorizing literature for easy reference and synthesis. Analytical Reading and Organization 3. Tools: 3. Tools: • ChatGPT and Perplexity : For generating the initial structure of your literature review. ChatGPT and Perplexity • PaperPal and Jenni AI : Tools to assist with proofreading your initial drafts and refining your literature structure. PaperPal and Jenni AI • Elicit or Scispace : To search for and summarize seed papers based on your structured queries. Elicit or Scispace • Litmap and Connected Papers : For discovering related literature reviews. Litmap and Connected Papers • Mendeley or Zetro : To manage and organize your references and PDFs. Mendeley or Zetro • Docanalyzer.AI : For analyzing documents and aligning them with your review structure. Docanalyzer.AI By aligning your research with this approach, you’ll streamline your workflow and lay a robust foundation for your literature review paper. Of course, this approach is flexible. For example, when you start structuring the plan for your literature review, you could use tools like Perplexity and then draft and proofread the initial paper structure using PaperPal. However, you could also gather information before you do any planning. Thus, a tactic like synthesizing information will need some techniques for targeted literature discovery through Elicit. So, please add or update any section of this approach that you feel fits your needs! structuring synthesizing information targeted literature discovery

AI Literature Reviews Are So Much Better With the Triple T Method

SINGAPORE, December 5, 2024 – Leading crypto exchange MEXC officially announced a comprehensive upgrade to its referral commission program, delivering enhanced benefits for users worldwide. Under the updated policy, referral commission rates for trading fees have increased to 40%, with select regions enjoying rates as high as 60%, setting a new benchmark for user incentives. As part of this upgrade, referrers can now earn up to 60 USDT in Futures bonuses for each valid invitee, while every valid invitee will receive a 20 USDT Futures bonus. This upgrade not only boosts referrer earnings but also ensures mutual benefits for both referrers and their invitees. With no increase in referral requirements, users can enjoy significantly higher rewards under the updated policy, reflecting MEXC's dedication to enhancing user benefits. By inviting friends, users can also access these additional perks: Trending Token Airdrops: Users can invite friends to participate in the Token Airdrop events and complete tasks to earn a share of token rewards from premium projects.
Reward Multipliers: Users can boost their earnings by inviting friends to participate in MEXC's Kickstarter events. Referring more valid invitees increases the reward multiplier, resulting in a larger share of token rewards. This upgrade underscores MEXC’s mission to empower its users while strengthening its community ecosystem. By enabling users to share in the platform's success, MEXC reaffirms its commitment to providing exceptional value. Looking ahead, MEXC will continue to innovate, enhance user experiences, and introduce more rewarding programs for its global community. About MEXC Founded in 2018, MEXC is committed to being "Your Easiest Way to Crypto." Serving over 30 million users across 170+ countries, MEXC is known for its broad selection of trending tokens, frequent airdrop opportunities, and low trading fees. Our user-friendly platform is designed to support both new traders and experienced investors, offering secure and efficient access to digital assets. MEXC prioritizes simplicity and innovation, making crypto trading more accessible and rewarding. Official Website｜X｜Telegram｜Invite Friends｜How to Invite Friends Disclaimer: Cryptocurrency investments carry inherent risks. The information provided does not constitute financial advice. It is important to fully understand the associated risks and proceed with caution when investing. Users are encouraged to conduct thorough research and assess their individual risk tolerance. This story was distributed as a release by Btcwire under HackerNoon’s Business Blogging Program. Learn more about the program here SINGAPORE, December 5, 2024 – Leading crypto exchange MEXC officially announced a comprehensive upgrade to its referral commission program, delivering enhanced benefits for users worldwide. Under the updated policy, referral commission rates for trading fees have increased to 40%, with select regions enjoying rates as high as 60%, setting a new benchmark for user incentives. As part of this upgrade, referrers can now earn up to 60 USDT in Futures bonuses for each valid invitee, while every valid invitee will receive a 20 USDT Futures bonus. This upgrade not only boosts referrer earnings but also ensures mutual benefits for both referrers and their invitees. With no increase in referral requirements, users can enjoy significantly higher rewards under the updated policy, reflecting MEXC's dedication to enhancing user benefits. By inviting friends, users can also access these additional perks: Trending Token Airdrops: Users can invite friends to participate in the Token Airdrop events and complete tasks to earn a share of token rewards from premium projects. Reward Multipliers: Users can boost their earnings by inviting friends to participate in MEXC's Kickstarter events. Referring more valid invitees increases the reward multiplier, resulting in a larger share of token rewards. Trending Token Airdrops: Users can invite friends to participate in the Token Airdrop events and complete tasks to earn a share of token rewards from premium projects. Token Airdrop Token Airdrop Reward Multipliers: Users can boost their earnings by inviting friends to participate in MEXC's Kickstarter events . Referring more valid invitees increases the reward multiplier, resulting in a larger share of token rewards. MEXC's Kickstarter events MEXC's Kickstarter events This upgrade underscores MEXC’s mission to empower its users while strengthening its community ecosystem. By enabling users to share in the platform's success, MEXC reaffirms its commitment to providing exceptional value. Looking ahead, MEXC will continue to innovate, enhance user experiences, and introduce more rewarding programs for its global community. About MEXC Founded in 2018, MEXC is committed to being "Your Easiest Way to Crypto." Serving over 30 million users across 170+ countries, MEXC is known for its broad selection of trending tokens, frequent airdrop opportunities, and low trading fees. Our user-friendly platform is designed to support both new traders and experienced investors, offering secure and efficient access to digital assets. MEXC prioritizes simplicity and innovation, making crypto trading more accessible and rewarding. Official Website｜ X ｜ Telegram ｜ Invite Friends ｜ How to Invite Friends Official Website ｜ Official Website X ｜ Telegram ｜ Invite Friends ｜ How to Invite Friends Disclaimer: Cryptocurrency investments carry inherent risks. The information provided does not constitute financial advice. It is important to fully understand the associated risks and proceed with caution when investing. Users are encouraged to conduct thorough research and assess their individual risk tolerance. This story was distributed as a release by Btcwire under HackerNoon’s Business Blogging Program. Learn more about the program here This story was distributed as a release by Btcwire under HackerNoon’s Business Blogging Program. Learn more about the program here here here

MEXC Referral Program Upgrade: Bigger Rewards, Better Benefits!

የMEXC ሪፈራል ፕሮግራም ማሻሻል፡ ትላልቅ ሽልማቶች፣ የተሻሉ ጥቅሞች!

MEXC Programa de Referencias ukan machaqar tukuyaña: ¡Juk’amp jach’a premios, juk’amp suma askinak!

Actualización del programa de referidos de MEXC: ¡mayores recompensas, mejores beneficios!

MEXC Erreferentzia-programaren bertsio berritzea: sari handiagoak, abantaila hobeak!

ارتقاء برنامه ارجاع MEXC: پاداش های بیشتر، مزایای بهتر!

Actualización do programa de referencia de MEXC: maiores recompensas, mellores beneficios!

MEXC 紹介プログラムのアップグレード: より大きな報酬、より良い特典!

Aktualizácia programu odporúčaní MEXC: väčšie odmeny, lepšie výhody!

MEXC Referral Program Uppgradering: Större belöningar, bättre förmåner!

อัปเกรดโปรแกรมอ้างอิง MEXC: รางวัลที่มากขึ้น ผลประโยชน์ที่ดียิ่งขึ้น!

MEXC Yönlendirme Programı Yükseltmesi: Daha Büyük Ödüller, Daha İyi Avantajlar!

MEXC 推荐计划升级：奖励更大，福利更好！

Authors:
(1) Limeng Zhang, Centre for Research on Engineering Software Technologies (CREST), The University of Adelaide, Australia;
(2) M. Ali Babar, Centre for Research on Engineering Software Technologies (CREST), The University of Adelaide, Australia. Table of Links Abstract and 1 Introduction 1.1 Configuration Parameter Tuning Challenges and 1.2 Contributions 2 Tuning Objectives 3 Overview of Tuning Framework 4 Workload Characterization and 4.1 Query-level Characterization 4.2 Runtime-based Characterization 5 Feature Pruning and 5.1 Workload-level Pruning 5.2 Configuration-level Pruning 5.3 Summary 6 Knowledge from Experience 7 Configuration Recommendation and 7.1 Bayesian Optimization 7.2 Neural Network 7.3 Reinforcement Learning 7.4 Search-based Solutions 8 Experimental Setting 9 Related Work 10 Discussion and Conclusion, and References 5 FEATURE PRUNING Given the diversity of workloads and the high-dimensional configurable parameters of DBMSs, both data collection and configuration space search can be time-intensive. Hence, employing strategies to mitigate workload execution time and streamline the search space becomes imperative. In this section, we present pruning strategies at two levels: workload-level pruning to optimize workload execution time and configuration-level pruning for enhancing search space efficiency. Table 2 provides a technical comparison of strategies on feature pruning (in Section 5) and knowledge extraction (in Section 6) from existing tuning methods. 5.1 Workload-level Pruning Workload-level pruning strategies primarily contribute to two aspects: data collection and model training. Query-level filtering aims to remove unimportant queries from a workload, thereby reducing the execution time of each workload and accelerating both training data collection and model validation. Workload Feature Selection aims to reduce the dimensionality of each workload feature, thereby decreasing the training time for autotuning models. 5.1.1 Query-level Filtering Considering that each application consists of a number of queries, the execution time of a workload can be decreased through removing some unimportant queries. Query filtering aims to removing the unimportant queries which are not sensitive to the configuration parameters. • Coefficient of Variation (CV): is a statistical measure that provides insights into the relative variability of a set of values. It is calculated as the ratio of the standard deviation to the mean, expressed as a percentage. LOCAT [7] utilizes CV to identify and eliminate configuration-insensitive queries whose performance (query execution time) is not affected by configuration changes within an application. It first constructs a matrix of the execution time for each query under different Spark SQL configurations (30 configuration instances) and then calculates the CV to filter queries with smaller CV values, categorizing them as ’configuration-insensitive’. 5.1.2 Workload Feature Selection Workload feature selection aims to reduce the dimensionality of workload features. Existing strategies applied in the automatic parameter tuning on DBMSs include: • k-means: is a popular unsupervised machine learning algorithm used for partitioning a dataset into clusters based on similarity. The primary goal of kmeans is to group data points in such a way that points within the same cluster are more similar to each other than to points in other clusters. In Ottertune [8], each workload is represented with runtime metrics. When conducting workload feature selection, factor analysis (FA) is first applied to transform the (potentially) high-dimensional DBMS metric data into lower-dimensional features. Then k-means, to cluster this lower dimensional data into meaningful clusters (for reducing noise data) and select a single metric for each cluster, namely, the one closest to the cluster center. During the process, number of clusters is determined through a heuristic model proposed in the work [25]. • Factor analysis (FA): is a statistical technique used for identifying underlying latent factors that explain the observed relationships among a set of variables. The key assumption in factor analysis is that observed variables are linear combinations of unobservable latent factors and random error terms. Ottertune [8] utilizes it to transform the (potentially) high dimensional DBMS metric embedding vector into lower-dimensional features. • Principal Component Analysis (PCA): is a popular linear dimension reduction method that transforms correlated variables into uncorrelated ones, known as principal components. It does so by calculating the covariance matrix, performing eigendecomposition, and selecting a subset of principal components based on their variance. HUNTER [13] applies PCA to complete the feature reduction on its collected workload runtime metrics for a workload, which including number of lush statements executed, number of threads activated, etc. • Autoencoder: is a neural network architecture meticulously devised to achieve precise reconstruction of its input. Comprising two integral components, namely the encoder and the decoder, it operates by first directing the input through the encoder to produce a latent code. This latent code subsequently serves as the foundation for the decoder to generate the output. Thus, the latent code assumes a pivotal role in the identification and disentanglement of various sources of variation inherent within the input data. Through the reduction of the latent code’s dimension, the autoencoder facilitates effective dimension reduction, thereby yielding a more condensed and meaningful representation of the input data. MMOT [22] utilizes the autoencoder architecture to extract inherent patterns among workloads and then uses these patterns to identify the model from similar workloads to generate configurations. This paper is available on arxiv under CC BY 4.0 DEED. Authors: (1) Limeng Zhang, Centre for Research on Engineering Software Technologies (CREST), The University of Adelaide, Australia; (2) M. Ali Babar, Centre for Research on Engineering Software Technologies (CREST), The University of Adelaide, Australia. Authors: Authors: (1) Limeng Zhang, Centre for Research on Engineering Software Technologies (CREST), The University of Adelaide, Australia; (2) M. Ali Babar, Centre for Research on Engineering Software Technologies (CREST), The University of Adelaide, Australia. Table of Links Abstract and 1 Introduction Abstract and 1 Introduction 1.1 Configuration Parameter Tuning Challenges and 1.2 Contributions 1.1 Configuration Parameter Tuning Challenges and 1.2 Contributions 2 Tuning Objectives 2 Tuning Objectives 3 Overview of Tuning Framework 3 Overview of Tuning Framework 4 Workload Characterization and 4.1 Query-level Characterization 4 Workload Characterization and 4.1 Query-level Characterization 4.2 Runtime-based Characterization 4.2 Runtime-based Characterization 5 Feature Pruning and 5.1 Workload-level Pruning 5 Feature Pruning and 5.1 Workload-level Pruning 5.2 Configuration-level Pruning 5.2 Configuration-level Pruning 5.3 Summary 5.3 Summary 6 Knowledge from Experience 6 Knowledge from Experience 7 Configuration Recommendation and 7.1 Bayesian Optimization 7 Configuration Recommendation and 7.1 Bayesian Optimization 7.2 Neural Network 7.2 Neural Network 7.3 Reinforcement Learning 7.3 Reinforcement Learning 7.4 Search-based Solutions 7.4 Search-based Solutions 8 Experimental Setting 8 Experimental Setting 9 Related Work 9 Related Work 10 Discussion and Conclusion, and References 10 Discussion and Conclusion, and References 5 FEATURE PRUNING Given the diversity of workloads and the high-dimensional configurable parameters of DBMSs, both data collection and configuration space search can be time-intensive. Hence, employing strategies to mitigate workload execution time and streamline the search space becomes imperative. In this section, we present pruning strategies at two levels: workload-level pruning to optimize workload execution time and configuration-level pruning for enhancing search space efficiency. Table 2 provides a technical comparison of strategies on feature pruning (in Section 5) and knowledge extraction (in Section 6) from existing tuning methods. 5.1 Workload-level Pruning Workload-level pruning strategies primarily contribute to two aspects: data collection and model training. Query-level filtering aims to remove unimportant queries from a workload, thereby reducing the execution time of each workload and accelerating both training data collection and model validation. Workload Feature Selection aims to reduce the dimensionality of each workload feature, thereby decreasing the training time for autotuning models. 5.1.1 Query-level Filtering 5.1.1 Query-level Filtering Considering that each application consists of a number of queries, the execution time of a workload can be decreased through removing some unimportant queries. Query filtering aims to removing the unimportant queries which are not sensitive to the configuration parameters. • Coefficient of Variation (CV): is a statistical measure that provides insights into the relative variability of a set of values. It is calculated as the ratio of the standard deviation to the mean, expressed as a percentage. LOCAT [7] utilizes CV to identify and eliminate configuration-insensitive queries whose performance (query execution time) is not affected by configuration changes within an application. It first constructs a matrix of the execution time for each query under different Spark SQL configurations (30 configuration instances) and then calculates the CV to filter queries with smaller CV values, categorizing them as ’configuration-insensitive’. 5.1.2 Workload Feature Selection 5.1.2 Workload Feature Selection Workload feature selection aims to reduce the dimensionality of workload features. Existing strategies applied in the automatic parameter tuning on DBMSs include: • k-means: is a popular unsupervised machine learning algorithm used for partitioning a dataset into clusters based on similarity. The primary goal of kmeans is to group data points in such a way that points within the same cluster are more similar to each other than to points in other clusters. In Ottertune [8], each workload is represented with runtime metrics. When conducting workload feature selection, factor analysis (FA) is first applied to transform the (potentially) high-dimensional DBMS metric data into lower-dimensional features. Then k-means, to cluster this lower dimensional data into meaningful clusters (for reducing noise data) and select a single metric for each cluster, namely, the one closest to the cluster center. During the process, number of clusters is determined through a heuristic model proposed in the work [25]. • Factor analysis (FA): is a statistical technique used for identifying underlying latent factors that explain the observed relationships among a set of variables. The key assumption in factor analysis is that observed variables are linear combinations of unobservable latent factors and random error terms. Ottertune [8] utilizes it to transform the (potentially) high dimensional DBMS metric embedding vector into lower-dimensional features. • Principal Component Analysis (PCA): is a popular linear dimension reduction method that transforms correlated variables into uncorrelated ones, known as principal components. It does so by calculating the covariance matrix, performing eigendecomposition, and selecting a subset of principal components based on their variance. HUNTER [13] applies PCA to complete the feature reduction on its collected workload runtime metrics for a workload, which including number of lush statements executed, number of threads activated, etc. • Autoencoder: is a neural network architecture meticulously devised to achieve precise reconstruction of its input. Comprising two integral components, namely the encoder and the decoder, it operates by first directing the input through the encoder to produce a latent code. This latent code subsequently serves as the foundation for the decoder to generate the output. Thus, the latent code assumes a pivotal role in the identification and disentanglement of various sources of variation inherent within the input data. Through the reduction of the latent code’s dimension, the autoencoder facilitates effective dimension reduction, thereby yielding a more condensed and meaningful representation of the input data. MMOT [22] utilizes the autoencoder architecture to extract inherent patterns among workloads and then uses these patterns to identify the model from similar workloads to generate configurations. This paper is available on arxiv under CC BY 4.0 DEED. This paper is available on arxiv under CC BY 4.0 DEED. available on arxiv

Prune Your Way to Better Database Management System Performance

Author:
(1) David M. Markowitz, Department of Communication, Michigan State University, East Lansing, MI 48824. Editor's note: This is part 7 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Table of Links Abstract
The Benefits of Simple Writing
The Current Work
Study 1a: Method
Study 1a: Results
Study 1b: Method
Study 1b: Results
Study 2: Method
Study 2: Results
General Discussion, Acknowledgements, and References Study 1b: Results Distributions of the comparisons in this study are reflected in Figure 1. Indeed, GPT significance statements were written in a simpler manner than PNAS significance statements for the simplicity index, Welch’s t(1492.1) = 11.55, p < .001, Cohen’s d = 0.58, 95% CI [0.47, 0.69]. Specifically, GPT significance statements (M = 75.53%, SD = 5.57%) contained more common words than PNAS significance statements (M = 69.84%, SD = 7.45%), Welch’s t(1478.7) = 17.31, p < .001, Cohen’s d = 0.87, 95% CI [0.76, 0.97]. GPT significance statements (M = 17.59, SD = 11.15) were also more readable than PNAS significance statements (M = 12.86, SD = 14.27), Welch’s t(1510) = 7.39, p < .001, Cohen’s d = 0.37, 95% CI [0.27, 0.47]. However, GPT significance statements (M = 92.73, SD = 6.89) had a statistically equivalent analytic style as PNAS significance statements (M = 92.32, SD = 7.48), Welch’s t(1587.7) = 1.16, p = .246, Cohen’s d = 0.06, 95% CI [-0.04, 0.16]. All results were maintained when comparing GPT significance statements to PNAS abstracts and PNAS significance statements as well. Alternative Explanations One possible explanation for the Study 1b results is that there are content differences across the PNAS and GPT texts explaining or impacting such differences across groups. This concern was addressed in two ways. First, PNAS has various sections that authors submit to, and LIWC has categories to approximate words associated with such sections. For example, the LIWC category for political speech would approximate papers submitted the Social Science section, specifically Political Sciences. Several linguistic covariates were therefore examined to account for content-related differences across GPT and PNAS texts. After including overall affect/emotion and cognition (to control for topics within the Psychological Sciences section of PNAS), political speech (to control for topics within the Political Science section of PNAS), and physical references to the multivariate models (to control for topics within the Biological Sciences section of PNAS), all results were maintained except for Analytic writing, where GPT texts were more analytic than PNAS texts, which is also consistent with prior work (42). Please see the online supplement for additional LIWC differences across these text types. Content effects were also evaluated in a bottom-up manner using the Meaning Extraction Method to measure dominant themes across the GPT and PNAS texts (43, 44). The evidence in the online supplement states there were 8 themes reliably extracted from the data, ranging from basic methodological and research information to gene expression and cancer science. Controlling for these themes, including the prior LIWC content dimensions, revealed consistent results as well (see supplement). Therefore, Study 1b evidence is robust to content. Altogether, human authors write simpler for lay audiences than for scientific audiences (Study 1a), but Study 1b demonstrated artificial intelligence and large language models can do so more effectively (e.g., the effect size differences between GPT significance statements and PNAS significance statements was larger than human in Study 1a). The findings thus far are correlational and therefore need causal evidence to demonstrate the impact of these effects on human perceptions. In Study 2, participants were randomly assigned to read a GPT significance statement or PNAS significance statement from pairs of texts that appeared in the previous studies. Participants made perceptions about the author (e.g., intelligence, credibility, trustworthiness), judged the complexity of each text, and they rated how much they believed the author of each text was human or artificial intelligence. Only perceptions of the author were made because prior work suggests people generally report consistent ratings when asked about both scientists and their science in similar studies (9). This paper is available on arxiv under CC BY 4.0 DEED license. Author: (1) David M. Markowitz, Department of Communication, Michigan State University, East Lansing, MI 48824. Author: Author: (1) David M. Markowitz, Department of Communication, Michigan State University, East Lansing, MI 48824. Editor's note: This is part 7 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Editor's note: This is part 7 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Editor's note: This is part 7 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Table of Links Abstract The Benefits of Simple Writing The Current Work Study 1a: Method Study 1a: Results Study 1b: Method Study 1b: Results Study 2: Method Study 2: Results General Discussion, Acknowledgements, and References Abstract Abstract The Benefits of Simple Writing The Benefits of Simple Writing The Current Work The Current Work Study 1a: Method Study 1a: Method Study 1a: Results Study 1a: Results Study 1b: Method Study 1b: Method Study 1b: Results Study 1b: Results Study 2: Method Study 2: Method Study 2: Results Study 2: Results General Discussion, Acknowledgements, and References General Discussion, Acknowledgements, and References Study 1b: Results Distributions of the comparisons in this study are reflected in Figure 1. Indeed, GPT significance statements were written in a simpler manner than PNAS significance statements for the simplicity index, Welch’s t(1492.1) = 11.55, p < .001, Cohen’s d = 0.58, 95% CI [0.47, 0.69]. Specifically, GPT significance statements (M = 75.53%, SD = 5.57%) contained more common words than PNAS significance statements (M = 69.84%, SD = 7.45%), Welch’s t(1478.7) = 17.31, p < .001, Cohen’s d = 0.87, 95% CI [0.76, 0.97]. GPT significance statements (M = 17.59, SD = 11.15) were also more readable than PNAS significance statements (M = 12.86, SD = 14.27), Welch’s t(1510) = 7.39, p < .001, Cohen’s d = 0.37, 95% CI [0.27, 0.47]. However, GPT significance statements (M = 92.73, SD = 6.89) had a statistically equivalent analytic style as PNAS significance statements (M = 92.32, SD = 7.48), Welch’s t(1587.7) = 1.16, p = .246, Cohen’s d = 0.06, 95% CI [-0.04, 0.16]. All results were maintained when comparing GPT significance statements to PNAS abstracts and PNAS significance statements as well. Alternative Explanations Alternative Explanations One possible explanation for the Study 1b results is that there are content differences across the PNAS and GPT texts explaining or impacting such differences across groups. This concern was addressed in two ways. First, PNAS has various sections that authors submit to, and LIWC has categories to approximate words associated with such sections. For example, the LIWC category for political speech would approximate papers submitted the Social Science section, specifically Political Sciences. Several linguistic covariates were therefore examined to account for content-related differences across GPT and PNAS texts. After including overall affect/emotion and cognition (to control for topics within the Psychological Sciences section of PNAS), political speech (to control for topics within the Political Science section of PNAS), and physical references to the multivariate models (to control for topics within the Biological Sciences section of PNAS), all results were maintained except for Analytic writing, where GPT texts were more analytic than PNAS texts, which is also consistent with prior work (42). Please see the online supplement for additional LIWC differences across these text types. Content effects were also evaluated in a bottom-up manner using the Meaning Extraction Method to measure dominant themes across the GPT and PNAS texts (43, 44). The evidence in the online supplement states there were 8 themes reliably extracted from the data, ranging from basic methodological and research information to gene expression and cancer science. Controlling for these themes, including the prior LIWC content dimensions, revealed consistent results as well (see supplement). Therefore, Study 1b evidence is robust to content. Altogether, human authors write simpler for lay audiences than for scientific audiences (Study 1a), but Study 1b demonstrated artificial intelligence and large language models can do so more effectively (e.g., the effect size differences between GPT significance statements and PNAS significance statements was larger than human in Study 1a). The findings thus far are correlational and therefore need causal evidence to demonstrate the impact of these effects on human perceptions. In Study 2, participants were randomly assigned to read a GPT significance statement or PNAS significance statement from pairs of texts that appeared in the previous studies. Participants made perceptions about the author (e.g., intelligence, credibility, trustworthiness), judged the complexity of each text, and they rated how much they believed the author of each text was human or artificial intelligence. Only perceptions of the author were made because prior work suggests people generally report consistent ratings when asked about both scientists and their science in similar studies (9). This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv

AI Crushes It at Simplicity: GPT-4 Writes Science Summaries Better Than the Pros

Author:
(1) David M. Markowitz, Department of Communication, Michigan State University, East Lansing, MI 48824. Editor's note: This is part 5 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Table of Links Abstract
The Benefits of Simple Writing
The Current Work
Study 1a: Method
Study 1a: Results
Study 1b: Method
Study 1b: Results
Study 2: Method
Study 2: Results
General Discussion, Acknowledgements, and References Study 1a: Results Descriptive statistics for each language dimension and intercorrelations are in Table 1. As expected, lay summaries were linguistically simpler than scientific summaries of the same article, Welch’s t(65793) = 40.62, p < .001, Cohen’s d = 0.31, 95% CI [0.29, 0.32].[2] At the item level of the simplicity index, lay summaries (M = 69.77%, SD = 7.14%) contained more common words than scientific summaries (M = 67.79%, SD = 6.60%), Welch’s t(68741) = 37.79, p < .001, Cohen’s d = 0.29, 95% CI [0.27, 0.30]. Lay summaries (M = 92.34, SD = 7.95) also had a simpler linguistic style than scientific summaries (M = 94.31, SD = 5.19), Welch’s t(59561) = - 38.52, p < .001, Cohen’s d = 0.29, 95% CI [0.28, 0.31]. Finally, lay summaries (M = 12.96, SD = 13.93) were more readable than scientific summaries as well (M = 12.49, SD = 12.46), Welch’s t(68320) = 4.67, p < .001, Cohen’s d = 0.036, 95% CI [0.02, 0.05]. Together, while lay summaries were indeed linguistically simpler than scientific summaries at PNAS, the effect sizes between such groups were quite small and it is therefore unclear if general readers would be able to recognize or appreciate such differences. Can lay summaries be written even simpler, using generative AI tools, to produce more substantive effect sizes while maintaining the core content of each text? In the next study, a random selection of abstracts was submitted to a popular large language model, GPT-4, and were given the same instructions as PNAS authors on how to construct a significance statement. This paper is available on arxiv under CC BY 4.0 DEED license. [2] 95% Confidence Intervals were bootstrapped with 5,000 replicates. Author: (1) David M. Markowitz, Department of Communication, Michigan State University, East Lansing, MI 48824. Author: Author: (1) David M. Markowitz, Department of Communication, Michigan State University, East Lansing, MI 48824. Editor's note: This is part 5 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Editor's note: This is part 5 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Editor's note: This is part 5 of 10 of a paper evaluating the effectiveness of using generative AI to simplify science communication and enhance public trust in science. The rest of the paper can be accessed via the table of links below. Table of Links Abstract The Benefits of Simple Writing The Current Work Study 1a: Method Study 1a: Results Study 1b: Method Study 1b: Results Study 2: Method Study 2: Results General Discussion, Acknowledgements, and References Abstract Abstract The Benefits of Simple Writing The Benefits of Simple Writing The Current Work The Current Work Study 1a: Method Study 1a: Method Study 1a: Results Study 1a: Results Study 1b: Method Study 1b: Method Study 1b: Results Study 1b: Results Study 2: Method Study 2: Method Study 2: Results Study 2: Results General Discussion, Acknowledgements, and References General Discussion, Acknowledgements, and References Study 1a: Results Descriptive statistics for each language dimension and intercorrelations are in Table 1. As expected, lay summaries were linguistically simpler than scientific summaries of the same article, Welch’s t(65793) = 40.62, p < .001, Cohen’s d = 0.31, 95% CI [0.29, 0.32].[2] At the item level of the simplicity index, lay summaries (M = 69.77%, SD = 7.14%) contained more common words than scientific summaries (M = 67.79%, SD = 6.60%), Welch’s t(68741) = 37.79, p < .001, Cohen’s d = 0.29, 95% CI [0.27, 0.30]. Lay summaries (M = 92.34, SD = 7.95) also had a simpler linguistic style than scientific summaries (M = 94.31, SD = 5.19), Welch’s t(59561) = - 38.52, p < .001, Cohen’s d = 0.29, 95% CI [0.28, 0.31]. Finally, lay summaries (M = 12.96, SD = 13.93) were more readable than scientific summaries as well (M = 12.49, SD = 12.46), Welch’s t(68320) = 4.67, p < .001, Cohen’s d = 0.036, 95% CI [0.02, 0.05]. Together, while lay summaries were indeed linguistically simpler than scientific summaries at PNAS, the effect sizes between such groups were quite small and it is therefore unclear if general readers would be able to recognize or appreciate such differences. Can lay summaries be written even simpler, using generative AI tools, to produce more substantive effect sizes while maintaining the core content of each text? In the next study, a random selection of abstracts was submitted to a popular large language model, GPT-4, and were given the same instructions as PNAS authors on how to construct a significance statement. This paper is available on arxiv under CC BY 4.0 DEED license. This paper is available on arxiv under CC BY 4.0 DEED license. available on arxiv [2] 95% Confidence Intervals were bootstrapped with 5,000 replicates.

Science Summaries Are Simpler, but Not by Much—Can AI Do Better?

Stories, News & Mentions about BETTER

Better’s Stories on HackerNoon