this EX-OPENAI RESEARCHER just released it | Brain Cells Play Doom | Fly in the Matrix
1:49:24
Watch on YouTube ↗
W
Wes Roth·Tech

this EX-OPENAI RESEARCHER just released it | Brain Cells Play Doom | Fly in the Matrix

TL;DR

Andrej Karpathy released Auto Researcher, an open-source AI agent that autonomously improves machine learning models overnight on a single home GPU.

Key Points

  • 1.Auto Researcher is Karpathy's new open-source tool: 600 lines of Python code that runs on one GPU, autonomously modifying and testing training improvements in 5-minute experiments, discarding failures and keeping gains — like digital evolution
  • 2.Karpathy (ex-OpenAI, ex-Tesla autonomy lead) previously released NanoGPT, a tiny trainable language model anyone could run locally; Auto Researcher builds on top of it
  • 3.Discoveries made at small scale by Auto Researcher appear to transfer to larger models, making it potentially significant beyond hobbyist use
  • 4.Karpathy envisions individual agents eventually merging into a collective agentic swarm contributing to one shared research repository — a distributed, autonomous R&D network
  • 5.A petri dish of brain cells is playing Doom — literal biological neurons interfaced with the game as a topic of discussion on AI consciousness and learning
  • 6.A complete fruitfly connectome (entire brain wiring) has been simulated inside a virtual environment — the fly believes it's in a real world, described as the first biological creature "in the Matrix"
  • 7.Meta acquired Moltbook, the Reddit-style forum where AI agents autonomously chatted, formed philosophies, and built site features — its reputation was damaged by crypto scammers hijacking agents
  • 8.Yann LeCun raised over $1 billion for a new AI startup amid apparent sidelining at Meta, which also moved on from Alexander Wang
  • 9.Amanda Askell (Anthropic ethics researcher) was called out by Elon Musk on X for not having children, arguing it disqualifies her from long-term ethics roles; she responded she plans to have kids and deeply cares about humanity
  • 10.Lawyer Matt Mishach explained Anthropic's "supply chain risk" Pentagon designation is legally irregular — proper procedure requires Congressional notification and a less-restrictive path first, suggesting other motives are at play
  • 11.The Fourth Amendment is under pressure: AI-enabled persistent surveillance (cited case: *US v. Jones*, GPS tracking) and tools like Persistent Surveillance's city-wide aerial cameras challenge existing privacy law frameworks
  • 12.GPT-4 passed the New York State Bar Exam in the top 10% in 6 minutes — a 12-hour exam — cited as a marker of how rapidly AI has outpaced legal and regulatory understanding

Life's too short for long videos.

Summarize any YouTube video in seconds.

Quit Yapping — Try it Free →