Global AI Weekly

Issue number: 100 | Tuesday, June 3, 2025

Highlights

Global AI Weekly Hits 100 Editions!

Global AI Weekly Hits 100 Editions!

We’re thrilled to celebrate the 100th edition of the Global AI newsletter! Over the past year, our community has nearly doubled in size, and we’re more excited than ever to keep delivering the latest in AI research, trends, and insights straight to your inbox. A huge thank-you to our readers — whether you’ve been with us from the start or just joined. We’re also growing on YouTube, so definitely take a peek there as well. Here’s to 100 issues — and many more to come! Subscribe, share, and stay curious.

Global AI community

Google quietly released an app that lets you download and run AI models locally

Google quietly released an app that lets you download and run AI models locally

Google has introduced a new app called Google AI Edge Gallery, allowing users to run a variety of AI models locally on their smartphones. The app leverages models from the popular AI development platform Hugging Face, enabling more direct and efficient use of AI without relying on cloud-based processing. This move highlights Google's push toward making AI tools more accessible and functional on everyday devices.

techcrunch.com

DeepSeek's latest AI model a ‘big step backwards’ for free speech

DeepSeek's latest AI model a ‘big step backwards’ for free speech

DeepSeek's new AI model, R1 0528, has sparked concern among users due to tighter restrictions on acceptable topics of discussion. Many see this as a setback for free speech, pointing out that the increased limitations could stifle open dialogue and expression. Critics argue that such regressions may outweigh advancements in other areas of the technology.

artificialintelligence-news.com

Research

DeepTheorem: Advancing LLM Reasoning for Theorem Proving Through Natural  Language and Reinforcement Learning

DeepTheorem: Advancing LLM Reasoning for Theorem Proving Through Natural Language and Reinforcement Learning

Explore how DeepTheorem combines natural language processing and reinforcement learning to enhance large language model reasoning for theorem proving. This approach bridges formal logic and natural language, making complex proofs more accessible and intuitive. Engage in the conversation about advancing AI's capabilities in mathematical reasoning and problem-solving.

huggingface.co

s3: You Don't Need That Much Data to Train a Search Agent via RL

s3: You Don't Need That Much Data to Train a Search Agent via RL

This paper highlights a fresh perspective on training search agents through reinforcement learning, challenging the common assumption that massive datasets are necessary. It explores strategies to achieve effective training with significantly less data, offering insights that could streamline and simplify the process. A must-read for those interested in efficient machine learning techniques and advancing search agent performance.

huggingface.co

Video

Humans in AI - Zoe Wilson

Humans in AI - Zoe Wilson

This episode of Humans in AI features Zoe Wilson, a dynamic leader driving business campaigns at Accenture and Abernard. She shares her perspective on how AI is transforming organizations, society, and the way we approach work, encouraging us to embrace innovative thinking. Tune in to discover how AI is shaping a more connected and forward-looking future in this inspiring conversation.

youtube.com

Building agents in Copilot Studio using Model Context Protocol open standards and more

Building agents in Copilot Studio using Model Context Protocol open standards and more

This session from Microsoft Build 2025 explores how to create advanced agents in Copilot Studio by leveraging Azure AI Foundry services. Attendees will learn about using Azure AI Search, the Azure model catalog, and the Model-Context Protocol (MCP) to enhance agent capabilities. Key highlights include discussions on open standards, live demonstrations of the Zapier MCP setup, and an overview of management plane and agent functionalities. The session also emphasizes Microsoft's dedication to supporting open standards for innovative development.

youtube.com

Articles

The people who think AI might become conscious

The people who think AI might become conscious

As advancements in large language models continue to push boundaries, a growing group of experts is exploring the possibility of AI gaining consciousness. They weigh the ethical, philosophical, and technical implications of machines that could potentially think and feel. While perspectives vary, the discussion highlights both the awe-inspiring potential and the challenges of navigating this frontier.

bbc.com

Optimizing Assembly Code with LLMs: Reinforcement Learning Outperforms Traditional Compilers

Optimizing Assembly Code with LLMs: Reinforcement Learning Outperforms Traditional Compilers

This content explores how reinforcement learning, combined with large language models (LLMs), can optimize assembly code more effectively than traditional compilers. It highlights breakthroughs in improving code efficiency and performance by leveraging AI-driven techniques. This approach showcases the potential for AI to transform programming and software development processes.

marktechpost.com

Upcoming Events

LlamaCon | AI Innovation Event

LlamaCon | AI Innovation Event

LlamaCon is a one-day virtual event featuring insightful keynotes from industry leaders Mark Zuckerberg, Ali Ghodsi, and Satya Nadella. The event focuses on advancements in AI innovation, offering discussions on technological breakthroughs and the future of artificial intelligence. It’s an opportunity to hear from top innovators shaping the AI landscape.

llama.com

June: AgentCon 2025

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.

agentcon.dev

Code

GitHub - resemble-ai/chatterbox: SoTA open-source TTS

GitHub - resemble-ai/chatterbox: SoTA open-source TTS

Discover Chatterbox, a state-of-the-art open-source text-to-speech (TTS) system developed by resemble-ai. This project invites contributions from developers to enhance its features and capabilities. Join the community on GitHub to collaborate and work on advancing TTS technology.

github.com

GitHub - microsoft/mcp-for-beginners

GitHub - microsoft/mcp-for-beginners

This open-source curriculum introduces the Model Context Protocol (MCP) with a focus on its core concepts and fundamentals. It includes practical examples in popular programming languages like .NET, Java, TypeScript, JavaScript, and Python, making it accessible for developers of various backgrounds. Perfect for beginners, it provides hands-on learning to build a solid understanding of MCP.

github.com

Podcast

How I AI

How I AI

How I AI is a podcast hosted by Claire Vo, aimed at helping listeners harness AI tools to enhance their work and life. Each 30-minute episode features guests sharing practical ways they use AI, complete with live screen sharing and actionable tips you can implement right away. Whether you're looking to boost productivity or gain new skills, this podcast simplifies AI and makes it accessible for everyone.

open.spotify.com

>