OpenAI's ambitious plan: AI to autonomously conduct research by 2028
OpenAI just revealed a bold plan: over the coming years, it hopes to launch an AI capable of autonomously running research projects or operating without moment-to-moment human supervision, while humans remain in charge of goal-setting and oversight.
Announced by CEO Sam Altman and Chief Scientist Jakub Pachocki, the project kicks off with an "AI research intern," which the company says is the first step, designed to run massive experiments using huge computing power.
AI could independently handle entire research projects
If OpenAI pulls this off, their AI could come up with ideas, design experiments, and produce research reports or handle end-to-end research on well-defined problems with limited human supervision.
The company's putting roughly $1.4 trillion to build out computing infrastructure (around 30 gigawatts) to support that work.
A 5-part safety system to ensure smooth transition
OpenAI says it's serious about keeping things safe.
They're rolling out a five-part safety system, making sure the AI's values and goals are in line with ours, checking for reliability and security risks, and monitoring how the AI thinks using something called "Chain-of-Thought Faithfulness."
The idea is to spot issues early and keep things transparent as the tech evolves.