I’ve got 7 months to go on my PhD here in Switzerland and so I’m rather busy — too busy to keep up with blogging for the moment. Currently I’m going through about 20 pages of hand written learning theory equations that Marcus Hutter derived. These are just raw equations with the occasional word or two in English, so it took me a while to come up to speed with the semantics of what was going on. Amazingly, I have yet to find one mistake in his workings. There are a few places however were he gets a bit bogged down, so perhaps I can find some new extensions or fill in some gaps. Aside from this work, I’ve also got some work on approximating AIXI and then of course I’ve got to put my thesis together. Like I said, not much time for blogging.
Another thing that I need to do is to figure out what I’m going to do next. I know two AI theory guys and both of them have been looking for permanent academic positions over the last six months. Neither of them have succeeded, even though they are both extremely good at what they do. This has given me a bit of a fright. I’m almost 33 now, and so by the time I’ve completed two post docs I’ll be 37. If I have trouble finding a permanent job at that point, being forced to switch back to industry again would not be easy. If I’m going to switch, it might be better to do so sooner rather than later. I’m thinking that this might be my best option. While the commercial world has limitations, unless you’ve got a permanent position academia can be just as limiting.
But what will I do in business? Obviously something involving AI and computers. Maybe bio-informatics? Market prediction? Something internet related? Work for Google? My first degree was in mathematical economics, so perhaps something in that area. Ideally I’d like to do something where I could develop AGI relevant technologies, given that that’s ultimately what I’d like to do.
… I’ll have to mull over this during the coming months while I finish my PhD.
There’s a job going over at Eliezer’s Singularity Institute, you know. The work is as conceptually intense as you could possibly hope for, but it would require enormous commitment – “give me Singularity or give me death”. I guess you have seven months to think it over.
The thing is, I’m rather skeptical about all this.
My biggest problem, as I commented in my “FAI is bunk” post, is that I just can’t get my head around what “friendly” means post singularity. What I suspect is that our concept of “friendly” is actually part of our primate brain and our culture. If we were from a different culture our sense of ethics would be different. If we were from a different species it could be very very different. Thus ethics might be as meaningful on a universal scale as, say, the concept of “good food”. Worms are nice to eat if you’re a bird, not if you’re a human. Similarly, it’s ok to eat your children if you come from certain species of lizard, but not if you’re a human. However, if ethics aren’t universal, exactly what are we trying to achieve? Are we trying to project human ethics into a future full of profoundly inhuman entities? Even our most fundamental ethical concepts like death and pain may be meaningless to them.
Good point Mitchell. We are indeed looking for another mathy person to do some hard mental crunching on all this. Actually if our current fundraising attempts succeed we will be looking for several people. This includes attempts to “break things” and to develop proofs showing how things will fail.
I think Eliezer’s current thinking on the issue is “let the FAI figure it out”. Give it a goal to figure out what humanity really would want, especially after it grows up further, and implement that. What it decides it wants to implement may very well allow for further changes in the future as cultures change, or it could provide a system where cultures can live under complete different rules. Who knows at this point.
On the “academia versus industry” theme — I note that the computer science academic job market is fairly robust, even if not in the particular domain of “formal AGI theory” that is the province of your thesis research. And, you are a diverse guy with interests and expertise beyond your thesis area. So, one option would be for you to get a postdoc, then spend this time period getting a few publications in other, more practical areas of computer science, as well as doing further research in formal AGI theory. This would position you well to get academic CS positions. For instance bioinformatics is quite “hot” these days in academia and is not a terribly difficult area to break into if you are an interdisciplinary thinker.
Another comment is that it is relatively easy to move from academia into industry at any stage, if you have any practical CS skills at all (which I know you do). OTOH it is not very easy to move from industry back into academia, unless during your period in industry you have bothered to keep churning out publications.
If you want to keep working on AGI I would say it makes sense for you to play the academic game a bit, generate a more diverse publication record, and get an academic CS job — or else go for some oddball job (no insult intended!!) like the SingInst one.
OTOH, if you want to turn your math and CS expertise into a pile of money, and are willing to work very hard in a not-so-entertaining environment for a significant number of years in order to achieve this, quantitative finance is definitely a realistic option…
– Ben G
Currently it looks like I’ll be doing a post doc in Switzerland using artificial intelligence to study financial models. That’s a half time job. With the rest of my time I’ll be doing some commercial applied machine learning work. Part of the reason for deciding on this mix was that I wanted to stay in Switzerland because I had a Swiss girlfriend. However we’ve now broken up so…
I didn’t think it would be easy going from academic AI to industry after 35 years of age. And acadmic AI does bug me a bit. Everything has to be so “serious” all the time. You can’t just cut loose and try something radical without risking your job. You spend half your time trying to convince people that what you’re wanting to do is not radical and yet still new.
Perhaps in the end the only way will be for me to try to start my own AGI company…?
Don´t you dear! Please join forces with SIAI, Eliezer, Ben and the others! AGI is serious business and I think we really need all critical brain mass to work together now, don´t let human social factors come in between!