Jason (jcreed) wrote,
Jason
jcreed

Went to a talk by Sungwoo about his research, and am now sitting in a talk by Marvin Minksy. I think I'm in an impatient mood, because both of them are feeling excruciatingly slow. Also Sungwoo's formalism had some glaring holes early on, so I found it hard to understand the actual novelty of his work, which followed it considerably later.

Edited to add:
Things I have learned from Marvin Minsky:

* that The General Problem Solver of Newell and Simon is the best idea in AI, ever. Furthermore, representing the mind with production rules is the worst idea in AI, ever.

* emotions and "rational thinking" don't seem subjectively different. Apparently, if you think they do, you're wrong.

* one of the dumbest things to happen in science is Chomsky arguing that grammar (Minsky's choice of word - I would have chosen "syntax" if I were talking about Chomsky, but...) is the most important aspect of language.

* Neuroscience is mostly irrelevant

* Freud invented neural nets

* A three-layer cognitive model is ridiculous. A six-layer cognitive model is eminently plausible, evidently, by virtue of having enough layers.

* Occam's Razor is total crap

* My theory isn't more complicated, it just has more boxes!

* Are you saying I should cut down my theory to the number of boxes I have evidence for???

* Are you saying six is more complicated than three???

* Lakoff: Brilliant

* Legitimate questions about falsifiability/verifiability of his particular theories are best answered by filibustering irrelevant facts about Popper being a fuckup, not that I disagree.

* Legitimate followup questions of the form, "no, we're concerned about the testability of your theory of consciousness" are best dodged by saying that we shouldn't be talking about consciousness at all, despite the fact that we as the audience have no better label on his theory other than as a theory of consciousness, or perhaps intelligence, or some other (admittedly vague and underdetermined!) term that he could pull the same move on.


* Computers will be pretty smart in: two years.

Man, this is the most boring GOFAI talk ever. Although I do agree with him that finding monosemy probably means that you're talking about mathematics --- then again, I haven't been able to articulate this thought clearly, and I think it might be very important to do so.
Subscribe
  • Post a new comment

    Error

    Anonymous comments are disabled in this journal

    default userpic

    Your reply will be screened

    Your IP address will be recorded 

  • 16 comments