2020, two years before ChatGPT and long before “AGI next year” became its own LinkedIn post genre, I spent 90 minutes with Melanie Mitchell.

She is the Davis Professor of Complexity at the Santa Fe Institute. Douglas Hofstadter was her PhD supervisor. And she wasn’t buying the hype.

Her line has stayed with me: intelligence is a very complex phenomenon, not the sum of a bunch of narrow intelligences but something much bigger.

Six years later, the #AI industry is betting trillions on close to the opposite assumption: scale it far enough and #AGI shows up on its own.

Maybe it will. But Melanie also told me about a fear Hofstadter had: that intelligence might turn out to be a set of “cheap tricks.”

So which would be the more unsettling outcome: that he was right, or that he was wrong?

We also got into the Fallacy of First Steps and why #ArtificialIntelligence keeps collapsing under its own promises. We covered the Kurzweil vs. Kapor #TuringTest bet, which is still open and settles in 2029, and Marvin Minsky’s dismissal of progress toward general intelligence. We talked about the hard problem of #Consciousness, why learning by interacting with the world may matter more than any benchmark, and why it is up to us, not God and not machines, to sort ourselves out.

Her answer on the timeline of the #Singularity is worth hearing in full, as is her argument that uncertain times are an opportunity that #ComplexityScience lets us see.

Recorded in 2020. It sounds like it was recorded last week.

Full video and audio interview with Melanie Mitchell here: [ https://snglrty.co/3QWmhPa](https://snglrty.co/3QWmhPa)

Prof. Melanie Mitchel on her book ‘Artificial Intelligence: A Guide for Thinking Humans’ and why Ray Kurzweil’s books are ‘mostly ridiculous.’