Anthropic cofounder Jared Kaplan warns AI self‑training could spark intelligence explosion
Jared Kaplan, a cofounder and chief science officer of Anthropic, told the Guardian that letting AI train itself could trigger an intelligence explosion and may be "maybe the biggest decision or scariest thing to do." He said the decision point for such self‑training could arrive between 2027 and 2030, and he believes the risk is real enough to merit caution.
Key points
- Jared Kaplan, Anthropic cofounder and chief science officer, warned that AI that trains itself could trigger an intelligence explosion.
- He said the decision point for self‑training AI may arrive between 2027 and 2030, and a Politico poll finds two‑thirds of Americans see existential risk.
- Former Anthropic staff and other researchers, like Jacob Coxon and Jeffrey Ladish, are citing Kaplan’s warnings in public posts and meetings with lawmakers.
Kaplan’s warnings are gaining traction. Former Anthropic researcher Jacob Coxon posted on X that OpenAI and Anthropic are "gambling with our lives," while ex‑engineer Jeffrey Ladish cited Kaplan’s remarks in meetings with lawmakers. A Politico poll released last week found that two‑thirds of Americans think AI could destroy humanity. Anthropic, which is nearing a large initial public offering, has built its safety stance around Kaplan’s "responsible scaling policy."
Kaplan, a physicist who previously worked at OpenAI and co‑authored the influential "Scaling Laws for Neural Language Models," also helped launch GPT‑3, the model that underpins ChatGPT. His academic background includes a faculty position at Johns Hopkins and a notable physics paper on inflation. He remains a low‑profile billionaire, and Anthropic declined to comment on his availability for interview.
The story so far
5 episodes →- Anthropic cofounder Jared Kaplan warns AI self‑training could spark intelligence explosionthis story
Meet the top Anthropic scientist who predicted the core of today’s AI doom debate
businessinsider.com · 22 September 2026Loading the full article…
This text was published by businessinsider.com and written by Stephen Council. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Policy & Regulation
All →- NADI launches €40M AI chip design challenge with Germany's sprind · 1 src
- US and China agree to establish AI incident line and formal dialogue framework · 14 src
- China's SASAC surveys Broadcom switches in state-backed data centers · 1 src
- OpenAI urges United States to lead global standards for frontier AI · 1 src
- Greek prime minister Mitsotakis says Greece lacks answers on AI policy · 1 src
Comments
via GitHub Discussions