#15301 · @BWEtradfi

*OPENAI CEO SAM ALTMAN SAYS HE DOESN'T THINK CO CAN PUSH SAFELY TOWARD RECURSIVE SELF-IMPROVEMENT NOW

中*OpenAI首席执行官Sam Altman表示,他认为公司目前无法安全地推进递归式自我改进

AI Overview

Updated Nov 16 19:39 UTC · 2 h after the alert

What changedNew addition: the arrangement for the company and Transluce to conduct a follow-up evaluation of about 8 weeks in early 2027

On September 6, 2026, ChatGPT developer OpenAI said it had achieved its "automated research intern" goal, capable of completing tasks that previously took researchers several days under human guidance.OpenAI

On November 16, the news account @tradfi relayed remarks from its CEO Sam Altman, saying he believes the company cannot yet safely pursue recursive self-improvement. This refers to AI continuously developing stronger successor models, progressively accelerating.OpenAI, @tradfi

On the same day, the company announced it had signed contracts with four independent institutions to conduct safety evaluations. The company and Transluce, one of them, plan to conduct another evaluation of about 8 weeks in early 2027. According to the company's previously announced plan, the next goal is to build a human-supervised automated AI researcher by March 2028.OpenAI

AI-generated from web searches. For reference only.