The Measure
Details
Warm Greetings to All,
Join Plato's Cave philosophers and Orlando Stoics on Zoom this Sunday morning, Aug. 16 at 9:00. (Informal chat at 9:00, forum at 9:15)
NOTE: New Applications for Plato's Cave membership are currently closed. The Plato's Cave forum will soon be discontinued on this meetup platform for reasons stated in my message to members.
Links and program information may be accessed through our partner organization, the Orlando Stoics meetup, at https://www.meetup.com/orlando-stoics/
Active Plato's Cave members can temporarily reserve a place, receive zoom login information on this site, and receive e-mail confirmation, until September 2026.
Every Sunday, a new forum. Our meeting starts at 9:00 AM with friendly wakeup chat; then our select topic panel briefly introduces the subject at 9:15, followed by member discussion and Q&A.
Volunteer introduction panelists will meet following each forum to develop the next program;.
This Sunday:
# The Measure
## Details
If intelligence is not one quantity, then "smarter than every human" does not mean anything definite. This week the group looks for a replacement. All three thinkers are working researchers rather than commentators, and each offers something testable.
We begin with Hans Moravec. In Mind Children (1988) he noted that it is easy to give a computer adult performance on an intelligence test or at checkers, and hard or impossible to give it the perception and mobility of a one year old. The reason is age. Chess and calculus are recent and shallow. Walking across a cluttered room draws on hundreds of millions of years of evolution, and because that work is finished and invisible to us, we count it as nothing.
We then turn to François Chollet, who treats intelligence as a rate of skill acquisition rather than a stock of skills. In "On the Measure of Intelligence" (2019) he argues that skill at a task measures prior knowledge and experience as much as anything else. A system that memorizes a domain and a system that can learn one score the same on ordinary benchmarks, which is why ARC uses puzzles that resist retrieval. The group’s live question now has an answer. ARC-AGI-1 fell. ARC-AGI-2 began at zero for every model in March 2025 and top models now score above 90 percent. ARC-AGI-3, which drops agents into unfamiliar interactive environments with no instructions, is still unbeaten at about 30 percent.
We arrive finally at Melanie Mitchell, who names the errors that produce confident timelines. In "Why AI Is Harder Than We Think" (2021) she sets out four fallacies. That narrow progress is a road to general competence. That easy things are easy, which is Moravec forgotten again. Wishful mnemonics, where we say a machine "understands" and then quietly believe the borrowed word. And that intelligence is computation in a brain rather than something a body maintains in a world. She traces all four to one source, which is that we are unaware of how complex our own thinking is.
Links
Hans Moravec
• Mind Children: The Future of Robot and Human Intelligence (1988, Harvard University Press): https://en.wikipedia.org/wiki/Mind_Children
• Moravec’s paradox, Wikipedia: [https://en.wikipedia.org/wiki/Moravec’s_paradox](https://en.wikipedia.org/wiki/Moravec%27s_paradox)
• "When Will Computer Hardware Match the Human Brain?" (1998, Journal of Evolution and Technology 1): https://jetpress.org/volume1/moravec.htm
• "Rise of the Robots" (1999, Scientific American 281:6, 124-135): https://frc.ri.cmu.edu/~hpm/project.archive/robot.papers/1999/SciAm.scan.html
• Hans Moravec, Wikipedia: https://en.wikipedia.org/wiki/Hans_Moravec
François Chollet
• "On the Measure of Intelligence" (2019): https://arxiv.org/abs/1911.01547
• "ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systems" (2025): https://arxiv.org/pdf/2505.11831
• ARC Prize leaderboard (check before the session, it moves): https://arcprize.org/leaderboard
• ARC-AGI-3, the interactive benchmark: https://arcprize.org/arc-agi/3
• ARC-AGI dataset and puzzle viewer: https://github.com/fchollet/ARC-AGI
• François Chollet, Wikipedia: [https://en.wikipedia.org/wiki/François_Chollet](https://en.wikipedia.org/wiki/Fran%C3%A7ois_Chollet)
Melanie Mitchell
• "Why AI Is Harder Than We Think" (2021): https://arxiv.org/abs/2104.12871
• Artificial Intelligence: A Guide for Thinking Humans (2019, Farrar, Straus and Giroux): https://en.wikipedia.org/wiki/Artificial_Intelligence:_A_Guide_for_Thinking_Humans
• Her homepage, Santa Fe Institute: https://melaniemitchell.me/
• AI: A Guide for Thinking Humans, her Substack: https://aiguide.substack.com/
• Melanie Mitchell, Wikipedia: https://en.wikipedia.org/wiki/Melanie_Mitchell
Hosted by:
Plato's Cave and Orlando Stoics
https://www.meetup.com/platoscave/
https://www.meetup.com/orlando-stoics/
=======================================================