OpenAI is forming a new team led by chief scientist and one of the company’s co-founders, Ilya Sutskever, to develop ways to manage and control “superintelligent” AI systems. This is reported by TechCrunch.
According to forecasts by Sutskever and the head of the company’s control group, Jan Leike, AI with intelligence exceeding that of humans could appear within a decade. This AI – assuming it ever does emerge – will not necessarily be benevolent, necessitating research into ways to control and limit it.
“Currently, we don’t have a solution for steering or controlling a potentially superintelligent AI, and preventing it from going rogue,” they write. “Our current techniques for aligning AI, such as reinforcement learning from human feedback, rely on humans’ ability to supervise AI. But humans won’t be able to reliably supervise AI systems much smarter than us.”
The new group will have access to 20% of the computing power that the company has provided to date. The team will work on solving the technical problems of controlling superintelligent AI over the next four years.
The group’s goal is to teach AI systems to use human feedback, to teach AI to help evaluate other AI systems, and eventually to create an AI that can conduct management research.
We will remind, according to the results of the research company Valoir, artificial intelligence has the potential to automate 40% of an average working day.