Home → Opinion → Articles → Insights for AI from the Human Mind → Abstract

Insights for AI from the Human Mind

By Gary Marcus, Ernest Davis

Communications of the ACM, Vol. 64 No. 1, Pages 38-41

[article image]

What magical trick makes us intelligent? The trick is that there is no trick. The power of intelligence stems from our vast diversity, not from any single, perfect principle.
—Marvin Minsky, The Society of Mind


Nandakumar Ramanathan

I'd like to bring to the notice of authors about a book entitled "Thinking, fast and slow" by Daniel Kahneman, Penguin Books, 2011 (ISBN: 978-1-846-14606-0). DK explains "the two systems that drive the way we think and make choices. One system is fast, intuitive and emotional; the other is slower, more deliberative and more logical." Both systems have their plus and minuses. I wonder how AI could be made to mimic human thinking in view of this dichotomy. R. Nandakumar (r_nand) aka Nandakumar Ramanathan.

Sathyanaraya Raghavachary

Nandakumar Ramanathan, the AI community is indeed aware of his work. He was a guest at a AAAI Fireside Chat event in April - https://vimeo.com/390814190, and last week, a speaker at #aidebate2: https://www.youtube.com/watch?v=VOI3Bb3p4GM

My own ("personal") view is this - there is no dichotomy, ie. no separate System1, System2 modes of processing for the brain to switch between; instead, it is a continuum, based on how much deliberate/conscious *attention* is needed in a situation. Familiar, routine (on account of past experience, practice...) situations are processed with very little attention, whereas novel, 'surprise', out-of-the-ordinary ones [for which there is no simple "lookup" based on the past] do need more attention.

Vinay Chaudhri

These insights reinforce the fundamentals and principles of AI that have been around for a long time. What is missing, however, is an articulation of a specific, measurable, research program that could be undertaken to make advances along one or more of these insights.

Displaying all 3 comments