Hacker News new | ask | show | jobs
Walter Pitts pioneered neural networks. Then he lit his entire PhD on fire (nautil.us)
73 points by bolmn 3104 days ago
4 comments

(early 2015)

previous discussion, 3 years ago, 23 comments: https://news.ycombinator.com/item?id=9003735

Really makes me wonder how such a brilliant person spends their time once they're not preoccupied with work. The article wasn't too clear on the precise timeline, but it sounded like there was over a decade where we're told he was just drinking—I'd be very curious to know more about the specifics of how he spent his time during that period. It may be that for the purposes of an article like this 'nothing of importance' took place, but I'd bet there are interesting aspects to it.
"There was a catch, though: This symbolic abstraction made the world transparent but the brain opaque. Once everything had been reduced to information governed by logic, the actual mechanics ceased to matter—the tradeoff for universal computation was ontology. ... anything and everything … can be done by an appropriate mechanism, and specifically by a neural mechanism—and that even one, definite mechanism can be ‘universal.’ Inverting the argument: Nothing that we may know or learn about the functioning of the organism can give, without ‘microscopic,’ cytological work any clues regarding the further details of the neural mechanism.”

- Would anyone mind to ELI5 this for me?

If something can do anything, than the way in which it can do _any_ thing doesn't much tell you how it does one _particular_ thing. In contrast, a mechanical clock can be inspected, and the way in which it tells time can be understood by its construction.
So, w.r.t to Pitt's research, proving that the brain could compute anything (like a Turing machine) does not help in understanding how it does one particular thing, for e.g. language?
Kind of. It's not the idea that the brain is a Turing machine that's the problem -- I think it's fairly well established that given enough scratch paper, a human _could_ manually execute an algorithm for any computable function. The problem is more specifically that in Pitt's model, all of the behavioral work is offloaded to the abstract logic, leaving none for the physical wetware. Of course, today we know that the brain does have quite a bit of physical structure, and we have learned a lot about how the brain functions by looking at these structures.
One can only speculate what he would have accomplished, had he lived a full life.