AI's own top researchers now think the technology could get scarily good at building better AI, fast.
In a paper published Monday, more than 20 AI researchers warned that AI systems capable of automating AI research could trigger what they call an 'intelligence explosion.' The signatories include AI pioneers Geoffrey Hinton and Yoshua Bengio, OpenAI chief scientist Jakub Pachocki, and Anthropic co-founder Jack Clark. Their central claim: once AI meaningfully speeds up its own research and development, progress that normally unfolds over years could instead happen within months. That such a diverse, high-profile group spanning competing labs and academia signed onto the same statement is itself worth noting.
A chief scientist at OpenAI and a co-founder at Anthropic - companies that compete fiercely and rarely align publicly - putting their names to the same warning suggests the risk is seen as credible by people closest to frontier AI development, not just outside critics. It also lands as those same labs keep building the automated coding and research tools that could plausibly set off the scenario they are warning about.
That tension - warning about a risk while continuing to build the tools that could cause it - says more about the incentives shaping AI right now than the letter itself does.