• Chaos Theory
  • Posts
  • 🥟 Chao-Down #269 Companies use AI to check employee messages, Researchers make an AI chatbot strong at answering chemistry questions, Patents on AI creations require "significant" human input

🥟 Chao-Down #269 Companies use AI to check employee messages, Researchers make an AI chatbot strong at answering chemistry questions, Patents on AI creations require "significant" human input

Plus, NVIDIA CEO Jensen Huang says countries must build their own sovereign AI infrastructure.

Does the industry really need $7 trillion(!) dollars to innovate in AI chips? The Nvidia CEO says no.

At the recent World Governments’ Summit in Dubai, Nvidia CEO Jensen Huang dismissed OpenAI CEO Sam Altman's plan to raise up to $7 trillion for an AI chip initiative, saying that the amount is unrealistic and ignores the advances in computer architecture. Huang said that Nvidia's CUDA architecture is unique in its ability to adapt to any AI challenge and is accessible to any researcher. He also said that Nvidia will have a $2 trillion install base of data centers by the end of the decade, powering software around the world. 

It remains to be seen whether Sam Altman will be successful in his ambitious fundraise. But can I just say that $7 trillion is still a LOT of money…

-Alex, your resident Chaos Coordinator.

What happened in AI? 📰

How Walmart, Delta & Starbucks are using AI to check employee messages (CNBC)

Patents on AI creations require 'significant' human input, USPTO says (Reuters)

AI chatbot shows surprising talent for predicting chemical properties and reactions (Nature)

AI Is Driving More Layoffs Than Companies Want to Admit (Bloomberg)

AI PCs will make up nearly 60% of total PC shipments by 2027 (TechSpot)

Nvidia CEO Huang says countries must build sovereign AI infrastructure (Reuters)

Always be Learnin’ 📕 📖

Large Language Models: A Survey (arxiv)

The data revolution in venture capital (signatureblock.co)

A Crash Course on Database Query Optimization Basics (SWE Quiz)

Projects to Keep an Eye On 🛠

OS-Copilot: Towards Generalist Computer Agents with Self-Improvement (Github)

Self-Play Fine-Tuning Converts Weak Language Models to Strong Language Models (Github)

noahfarr/rlx: A reinforcement learning framework based on MLX. (Github)

The Latest in AI Research 💡

Large Language Models are Superpositions of All Characters: Attaining Arbitrary Role-play via Self-Alignment (arxiv)

Efficient Tool Use with Chain-of-Abstraction Reasoning (arxiv)

AnyTool: Self-Reflective, Hierarchical Agents for Large-Scale API Calls (arxiv)

The World Outside of AI 🌎

The new car batteries that could power the electric vehicle revolution (Nature)

China claims new speed record with vacuum-tube maglev train (New Atlas)

Biden campaign joins TikTok in an effort to reach younger voters (The Washington Post)

Reddit Publishes Report Into How the Platform Is Influencing Purchase Decisions (Social Media Today)

Cassette tapes are making a surprise comeback in Japan (Nikkei Asia)

American CEOs Visiting China Can’t Escape It: They Have to Dance on Stage (WSJ)

One Last Bite 😋