I’ve got a backlog of things in my brain that I’ve been wanting to blog, but I’ve been rather busy of late. So here’s a somewhat random brain dump of what’s going on around here…
PhD thesis on AIXI, intelligence and all that is mostly written. I’ve got about one and a half chapters left to go and most of the difficult stuff is out of the way. Other than being an integrated work now rather than separate papers, the main new contribution is a chapter on AIXI that, I hope, is considerably easier and more enlightening than previous publications on this topic. Ironically, the better I do my job in making AIXI simple, the less impressed some people will be!
The thesis is unfortunately going a bit slow however. Once reason is that I’m now doing this post doc in finance. I now know how much finance I didn’t know. Or at least I have a better idea. I also have to keep on travelling to other parts of Switzerland to do courses which isn’t helping my thesis. I’m learning quite a bit about cutting edge asset pricing models however, which could be useful knowledge should I decide to leave the ivory tower and get into commercial applications of AI and machine learning.
Speaking of which, some recent events, that I won’t go into here, have got me thinking quite a bit about what direction I want to take after my post doc. I’m now thinking that the best way to do something really interesting in AI/AGI is to go commercial. Academia is just too limited. It’s not that I’m not doing well in this environment, indeed just in the last week I got a paper into Neural Information Processing Systems (NIPS, one of the very top AI conferences in the world), and also a 50 page journal paper on complexity and the nature of intelligence accepted. However I feel like I’m constantly struggling against a very conservative mindset in the field. If I could achieve a small amount of financial freedom though commercial applications of AI over the next five years, that would enable me to do some of the work that I think really needs to be done in order to move from narrow AI to full blown AGI systems. The timing perhaps isn’t too bad either as I don’t think the hardware that I’ll need for what I have in mind will be around for a few more years yet. Of course that isn’t going to stop me from playing with models and trying to solve some of the key design issues on a smaller scale before then.
Anyway, once I get this asset pricing stuff and my thesis and these various other papers cleaned up, expect to hear a bit more for me on my blog…
3 responses so far ↓
1 Ben Goertzel // Jan 9, 2008 at 6:25 pm
I agree that academia is very limiting in terms of AGI research — both in terms of the funding opportunities, and also the general cultural and intellectual climate.
On the other hand, industry is not generally any better as a home for AGI research.
I can surely see the appeal of spending N years getting rich or at least moderately well-off, and then spending one’s early retirement working on AGI.
And some folks have done just that: Jeff Hawkins and Georges Harik, to name two.
However, I’m sure a lot more folks have tried but then failed to actually accumulate enough wealth….
Perhaps I should have followed the path you suggest — but I’m glad I didn’t because I’ve learned so much and had so much fun working on AGI, in spite of the stress of doing it in parallel with money-generating activities.
2 Falafulu Fisi // Jan 9, 2008 at 9:33 pm
This is one of the journal that I frequently read , the Neural Information Processing Systems (NIPS). It is the best out there in the area of Computational Intelligence. What’s the title of your paper so I can look out for it to download?
3 mathemajician // Jan 9, 2008 at 9:46 pm
The paper is:
Temporal Difference Updating without a Learning Rate
Marcus Hutter and Shane Legg.
In Neural Information Processing Systems (NIPS ‘07)
Leave a Comment