Tim Rocktäschel, an AI professor at University College London, was teaching a class recently on a novel branch of research in his field when one of his students reached for a historical analogy to describe the work.
“We’re all kind of mini Oppenheimers at this point,” the student quipped. Rocktäschel, who until recently was a top researcher at Google DeepMind, had to concede it was a fair assessment.
The subject of the comparison to nuclear weapons research was the current race, by top AI labs including OpenAI and Anthropic and a growing group of start-ups, to develop an AI that can do something unprecedented in technology: build its own successor.
您已阅读4%（654字），剩余96%（14427字）包含更多重要信息，订阅以继续探索完整内容，并享受更多专属服务。