Jeff Hawkins at MIT Technology Review's EmTech Conference 2021: Building a Better AI

แชร์
ฝัง
  • เผยแพร่เมื่อ 11 พ.ย. 2024

ความคิดเห็น • 8

  • @szabolcsferenczi5558
    @szabolcsferenczi5558 3 ปีที่แล้ว +5

    20:20 "Why are we stuck on the idea that artificial intelligence must be human-like rather than animal-like..." It is not mentioned in this lecture, but I understand that Jeff Hawkins is interested in the neocortex, and not so much in the old brain. Elsewhere he says: "The seat of intelligence is the neocortex". All mammals have a neocortex. I believe the microstructure of the neocortex is the same in humans and other mammals, and both are intelligent. The difference is in the higher level structure, e.g. bears have the best sense of smell of all mammals, simply because bears have the largest olfactory bulb in the neocortex. On the other hand, I think humans have the evolutionary advantage of having evolved a large language area in the neocortex. However, I believe that language is only a catalyst for intelligence. (The exception is birds, which have a different brain structure, but they are still intelligent creatures.)

  • @2002budokan
    @2002budokan 3 ปีที่แล้ว +5

    What is the difference between Numenta's cortical columns and Hinton's capsules?

    • @NumentaTheory
      @NumentaTheory  3 ปีที่แล้ว +3

      We wrote a blog about the main differences between Numenta's Thousand Brains Theory and Hinton's recent GLOM model (The GLOM model builds on Hinton’s earlier work on Capsules). You can read it here: numenta.com/blog/2021/04/26/comparing-hinton-glom-model-to-numenta-thousand-brains-theory

    • @2002budokan
      @2002budokan 3 ปีที่แล้ว +1

      @@NumentaTheory Thank you, I will read it.

  • @chrism.1131
    @chrism.1131 2 ปีที่แล้ว

    What percentage of the brain is geared toward learning new stuff and what percentage remembers the stuff already learned?

  • @AlexeySlepov
    @AlexeySlepov 3 ปีที่แล้ว +1

    A capable to learn different things AI can be taught to do dangerous things like a terrorist act. How do you see an infrastructure to keep machines from learning bad things?

  • @jusme8060
    @jusme8060 3 ปีที่แล้ว

    Maybe we could use it to sift through all the pseudoscience and theories sold to us as facts. Like gravity for example. Or heliocentrism. Maybe it could go at germ theory or viruses and see what's really going on behind all that nonsense. Evolution would be a good one. I wonder if a.i would believe it evolved into existence :) Studying nature and copying God is definitely the way to make cool stuff though.