Global AI Weekly - Issue 102 - Global AI Community

In this issue

Highlights

2 stories

npr.org

In first-of-its-kind lawsuit, Hollywood giants sue AI firm for copyright infringement

Disney and Universal have filed a groundbreaking 110-page lawsuit against AI company Midjourney, alleging the unauthorized use of "countless" copyrighted works to train its software. The entertainment giants claim the AI player infringed on their intellectual property, marking a first-of-its-kind legal battle in the intersection of AI and copyright law. This case could set a major precedent for how AI companies handle copyrighted material in the future.

reuters.com

OpenAI taps Google in unprecedented cloud deal despite AI rivalry

In a surprising move, OpenAI has signed a cloud deal with Google to expand its computing capacity, easing its reliance on primary backer Microsoft. Finalized in May, the partnership highlights the immense infrastructure demands of AI development and reshapes the competitive landscape between the two tech giants. Google Cloud will now support OpenAI’s efforts to train and run its models, including those powering ChatGPT—Google’s biggest search rival. While all parties declined to comment, Alphabet’s stock rose 2.1% on the news. Analysts noted the deal’s strategic implications for Google Cloud, despite the ongoing competition between the companies in generative AI.

In this issue

Research

2 stories

arxiv.org

The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity

This paper explores the capabilities and boundaries of reasoning models by examining them through the perspective of problem complexity. It introduces a framework to assess how well these models handle various reasoning tasks and highlights situations where they may falter. The study provides insights into enhancing the design and evaluation of reasoning systems for improved performance and reliability.

huggingface.co

Confidence Is All You Need: Few-Shot RL Fine-Tuning of Language Models

This paper explores the concept of using confidence as the key ingredient for fine-tuning language models in few-shot reinforcement learning (RL) scenarios. It focuses on demonstrating how a confidence-driven approach can enhance the adaptability and performance of language models with minimal training examples. Practical insights and methodologies are shared to make fine-tuning more efficient and effective in RL contexts.

In this issue

Video

2 stories

youtube.com

Humans in AI - Ajit Jaokar

In this episode of "Silicon Minds, Human Hearts," Ajit Jaokar, an Oxford University professor, shares his passion for making AI accessible and impactful. He explores the evolving role of AI in education and creativity, reflecting on how it transforms professions and opens paths for human-AI collaboration. With thought-provoking examples and insights, Ajit highlights real-world successes and discusses how institutions like Oxford are shaping the future of AI empowerment.

youtube.com

What is Retrieval-Augmented Fine-Tuning (RAFT)?

Isaac Ke explains Retrieval-Augmented Fine-Tuning (RAFT), a cutting-edge approach combining Retrieval-Augmented Generation (RAG) and fine-tuning to tackle domain-specific AI tasks. RAFT enhances AI performance by reducing hallucinations, improving accuracy, and enabling chain-of-thought reasoning, making it a powerful tool for scaling enterprise AI. This innovative method bridges the gap between data retrieval and targeted model training, paving the way for more reliable and efficient AI solutions.

In this issue

Articles

3 stories

techcrunch.com

OpenAI releases o3-pro, a souped-up version of its o3 AI reasoning model

OpenAI has introduced o3-pro, its most advanced AI reasoning model to date, designed to tackle complex tasks with greater efficiency. The company highlights its improved capabilities over the previous o3 version, aiming to set a new benchmark in AI performance. This release reflects OpenAI's ongoing push to refine and expand its AI technologies.

zdnet.com

The best AI for coding in 2025...and what not to use

For over two years, David Gewirtz has been testing chatbots against real-world programming challenges to identify the best AI tools for coding. In 2025, there are five standout options he recommends for coding assistance, including a new top pick. He also highlights some AI tools that you should steer clear of to avoid wasting time and effort.

nature.com

Memory kernel minimization-based neural networks for discovering slow collective variables of biomolecular dynamics

In this issue

Upcoming Events

2 stories

agentcon.dev

June: AgentCon 2025

The AI Agents World Tour, part of AgentCon 2025, is a global series of one-day events tailored for developers working with autonomous AI agents. This June we will be bringing together top engineers and researchers. These conferences will take place in major cities like San Francisco and Singapore. It's a unique opportunity to connect, learn, and shape the future of AI technology.

$Scaling Development with Remote Agents: Best Practices and Deep Dive with Augment Code | Webinars \ Anthropic$

anthropic.com

Scaling Development with Remote Agents: Best Practices and Deep Dive with Augment Code | Webinars \ Anthropic

This webinar highlights strategies and tools for scaling development with remote agents, focusing on creating exceptional industry-leading experiences. It includes live demonstrations and actionable best practices designed to empower teams in enhancing agent performance effectively. Don't miss this opportunity to elevate your team's capabilities with practical insights.

In this issue

Code

2 stories

github.blog

Remote GitHub MCP Server is now in public preview

The Remote GitHub MCP Server is now available in public preview, making it easier to connect AI agents to GitHub tools and context. With features like OAuth integration, one-click setup, and automatic updates, this hosted server simplifies the process and enhances efficiency for users. Powered by GitHub's infrastructure, it streamlines workflows while maintaining ease of use.

GitHub - yamadashy/repomix: Repomix is a powerful tool that packs your entire repository into a single, AI-friendly file.

Repomix is a powerful tool that packs your entire repository into a single, AI-friendly file. It is perfect for when you need to feed your codebase to Large Language Models (LLMs) or other AI tools like Claude, ChatGPT, DeepSeek, Perplexity, Gemini, Gemma, Llama, Grok, and more.

In this issue

Podcast

1 story

open.spotify.com

Lex Fridman Podcast

The Lex Fridman Podcast features engaging conversations on topics like science, technology, history, philosophy, and human nature. Hosted by AI researcher Lex Fridman, the show explores themes such as intelligence, consciousness, love, and power, offering thought-provoking discussions with a wide range of experts and visionaries.