Human Minds Remain Special in Age of AI, Says Princeton Professor
In a Guardian opinion piece, Princeton professor Tom Griffiths argues that human intelligence remains distinct from AI due to biological constraints. He challenges the notion of intelligence as a single scale, comparing it to height. Griffiths notes that AI systems like AlphaGo and ChatGPT are trained on vast data, unlike humans who learn from limited experience. He cites examples where AI fails, such as GPT-4 miscounting letters and leading models choosing wrong drug concentrations. Human intelligence, shaped by finite lives and communication limits, excels in pattern recognition and collaboration. Griffiths concludes that AI and human minds will be different, not superior, and should be seen as companions rather than rivals.
Key facts
- Tom Griffiths is a professor of information technology at Princeton University.
- Griffiths authored 'The Laws of Thought' (William Collins).
- The article was published in The Guardian on May 3, 2026.
- Griffiths argues intelligence is not a single scale like height.
- AlphaGo was trained on many human lifetimes of games.
- ChatGPT draws on thousands of years of language data.
- GPT-4 was more likely to correctly answer a letter-count question for 30 letters than 29.
- AI systems sometimes pick 685 ppm over 791 ppm for drug concentration due to neural network blurring.
Entities
Artists
- Tom Griffiths
Institutions
- Princeton University
- The Guardian
- OpenAI
- William Collins
- Allen Lane
- Bodley Head
- Faber