2025-01-23

OpenAI’s Operator is designed to help users by executing tasks independently. It’s based on a new model called “computer-using agent” (CUA), which combines GPT-4o’s vision capabilities with advanced reasoning through reinforcement learning.

Operator is initially available to Pro users in the US, with plans to integrate it into ChatGPT for broader availability.

OpenAI emphasizes safety with Operator, having conducted extensive internal testing and external red teaming to ensure its secure usage.

Google DeepMind CEO Demis Hassabis announced that the Nobel Prize-winning AlphaFold team is now working on the Virtual Cell Project, aiming to simulate biological cells to aid in drug discovery within the next five years.

Demis Hassabis also mentioned that AI agents are on the horizon, capable of exploring the Tree of Knowledge to discover new insights about the world.

Google DeepMind anticipates AGI that can handle all cognitive tasks and generate new scientific hypotheses to be 3-5 years away.

@emollick discusses the implications of OpenAI’s Operator, noting that while general-purpose web agents are not fully realized, Operator shows promise. He also highlights the need for companies to consider marketing to AI agents and the upcoming security challenges.

@MatthewBerman shares that OpenAI’s Operator can control a browser and execute real-world tasks, expressing excitement about testing its capabilities.