Jacob Coxon discusses AI risks and possible return to a leading lab
At a Stanford University discussion, former OpenAI and Anthropic researcher Jacob Coxon said leaving a leading AI laboratory had left him anxious about missing new advances and suggested returning was possible. He also discussed super-exponential pretraining efficiency, recursive self-improvement and reinforcement-learning risks, while defending his decision to retain OpenAI stock because he liked the money. The technical claims, including assertions about model deception and an OpenAI sandbox incident, were personal statements reported by 36氪人工智能 without independent corroboration.
Coxon’s comments offer a notable former frontier-lab researcher’s perspective on pretraining progress and reinforcement-learning safety, but the reported technical claims and possible return remain unverified.
刚刚,两个月前高调裸辞的Anthropic前研究员Jacob Coxon,态度突然大反转了!
在斯坦福大学最新公开的对谈中,这位27岁的剑桥数学系才子当众承认——
自己不确定当初辞职到底对不对,甚至动了重返顶尖AI实验室的念头。
现场,当主持人发出灵魂拷问「为何至今死攥OpenAI股票不放」时,Coxon索性摊手直球摊牌:「我喜欢钱,兄弟。我承认这是带血的钱。」
果然,没有人能逃过钱的诱惑,哪怕是天天喊着「AI毁灭世界」的硬核末日论者也不例外!
#Jacob Coxon#OpenAI#Anthropic#Claude#Hugging Face#Stanford HAI