[<< | Prev | Index | Next | >>]

Sunday, November 05, 2000

Neurillogic

An excerpt of an email from a friend of mine who's a renowned professor of functional neurobiology, which sadly confirms what I've been saying all along but have hoped was simply pessimism:

Most work currently going on in "neural networks" per se does not seem very interesting to me these days -- things like minor tweaks to backprop, and the occasional attempt to say something about biological mechanisms, issued by people who have never looked in a microscope and wouldn't know a dendrite if it were floating in their martini. As fields grow larger, they get socialized too rapidly, and the social structures dictate what gets worked on, instead of people actually looking at real problems, real objects of study, and real goals. But hey; that's all the more space they leave for us! :-) Of course, those who are still outside the heavily socialized domains are pretty idiosyncratic (guilty as charged), so there's a broad spread of different work going on. There are some projects out there that are interesting to me, and perhaps they would be to you as well; tell me what kinds of things you want to do, and perhaps I'll think of someone who might be a useful collaborator.

Some good leads coming from that already....

He also related my plans for the next decade of my life to the first ten years of his own career. I asked him whether that was good or bad (grin)...:

It really was a good 10 years; I wouldn't trade it. You're not too old at all -- it should work just fine in reverse order -- the idea is to look for fun things that your Finely Honed Instincts tell you are bound to be worth studying even though you can't tell what they're good for ... yet. (I guess the catch is that, after you do this, you'll probably want to re-enter the market to show all the academic idiots that what you were working on during your "retirement" really was real, and really does have application, unlike what they're doing, and despite what they claim. :-)

You know it. :)

[<< | Prev | Index | Next | >>]