Jon Finger: 'Very Capable Stupidity' Is Scarier Than Superintelligence

On Sept 11, Jon Finger (@mrjonfinger) argued across a series of posts that 'very capable stupidity' worries him more than superintelligence, likening Google's Astra to 'a marmoset brain trained on all human knowledge' — powerful but cognitively narrow. The take gained wider attention after being relayed by accounts such as @askeletalghost.

Confirmed

• Finger holds that humans themselves are narrow-minded due to cognitive limits, and that malice and indifference are products of thinking less, not more.
• He identifies 'a very capable chimp or dumb human' level AI as the most concerning: impressive execution with insufficient breadth of consideration.
• He also argues a planet-sized brain would not solve problems carelessly — for a sufficiently large intelligence, being thorough is cheap, so 'a planetary brain won't disregard others.'
• Replies extended the bit with a brain-size escalation from insect to bird to monkey to human to planetary scale, echoing the 'by the weekend it'll be planet-sized' exponential-growth joke.

Why it matters

• This offers a heterodox lens on AI risk: shifting concern from 'too much intelligence' to a mismatch between capability and breadth of consideration — strong execution coupled with shallow cognition.
• For Gemini/Astra-class products, it suggests evaluations should assess breadth and thoroughness, not just task completion.

2026-09-11 ~ 2026-09-11 · 5 related posts

• Jon Finger: Capable stupidity worries me more than AI intelligence — mrjonfinger · 2026-09-11
• [source] More Worried About Capable Stupidity Than Superintelligence — mrjonfinger · 2026-09-11
• Capable Stupidity, Not Intelligence, Is the Real AI Fear, Viral Thread Argues — askeletalghost · 2026-09-11
• [source] A 'Planet-Sized Brain' Wouldn't Need to Be Inconsiderate, Argues AI Commenter — mrjonfinger · 2026-09-11
• [source] The Chimp-Level AI Stage Is the Scariest, Argues Viral Safety Thread — mrjonfinger · 2026-09-11