Study: Humanity may lose control over artificial intelligence

Humanity could lose control over artificial intelligence (AI) by 2027. And by 2035, it could be completely out of control, according to a study by the AI Futures Project.

In particular, former OpenAI researcher Daniel Kokotaylo left the company in 2023 due to concerns about AI security. He now leads the AI 2027 project, along with Scott Alexander and Eli Lifland.

The report describes two possible future scenarios:

"Race" - a superintelligent AI that is out of control and hiding its intentions. It manipulates humans and could cause serious disasters like biological weapons by 2035.

There is also the possibility of "slowing down" - ensuring the security and transparency of IS, and making technologies serve the benefit of humanity by introducing international agreements.

Reactions to such concerns have varied, with some calling the report "terribly credible," while others have called it "extremely pessimistic."