?

Log in

What Bill Gates wants for Christmas

« previous entry | next entry »
Dec. 14th, 2006 | 10:38 pm

I just read this, sort of an interview with Bill Gates. At one point he was asked what he wants for Christmas. His answer was lectures from here. Me too.

I depart for New Zealand in a few days. Man I hate that flight. Anyway, it will be nice being back in NZ again. I'm also spending almost two weeks in Rarotonga (tropical island) before heading to Los Angeles for a couple of days, and then back to Switzerland. Once I arrive back I'll have just 4.5 months before my PhD is over.

I'm starting to work out what to do next. Of course what I'd really like to spend the next 5 years doing would be creating a radical new artificial intelligence system. After working in two companies trying to do that, and having spent the last 4 years learning more about AI for my PhD, I know how I'd go about doing it and I'm reasonably confident that I could succeed. However I have no idea how I could finance it. I estimate that I'd need $1.5 m to get to the point of proof of concept. It would require about 3 people, 10 computers and 5 years work. Once proof of concept had been achieved money would be no problem — the company that is first to develop real AI will easily have a market cap north of a trillion dollars. Alas I don't know any angel investors with that kind of money and outlandish ambition.

Plan B. Currently this plan looks like doing a post doc, perhaps a part time one, in finance. I'd be based in Switzerland and would apply machine learning methods to problems in finance research.

Link | Leave a comment | Share

Comments {9}

domino plural

(no subject)

from: spoonless
date: Dec. 15th, 2006 01:25 am (UTC)
Link

3 people, 10 computers, 5 years of work, and 1.5-million dollars, eh?

Is the $1.5million so that you can pay each of the 3 $100,000 per year, or is there additional cost requirements beyond getting smart people to work on the problem?

Reply | Thread

mathemajician

(no subject)

from: mathemajician
date: Dec. 15th, 2006 11:13 am (UTC)
Link

There are always additional costs beyond simply salary.

Computer hardware, computer software, office space, internet connection, power, getting visas, travel to conferences and conference fees, travel to see particular experts whose advice you might need, buying particular books, various other things that don't come to mind at the moment. It all adds up.

Also, 5 years is not a trivial amount of time during which to be earning on a low income and on a project like this staff turnover is bad news. Thus if you want good people who will stay the 5 years needed it's best to pay them a reasonable salary. Indeed if this was to be done in the US you have to pay them well in order to be able to get the visa approved — although I think the US would be a bad idea because the visa system post 9/11 is a big problem (the H1-B visa cap is hit of the first day of each financial year). It would be much easier doing it in other countries like Canada, Australia, New Zealand, Switzerland that are more open to bringing in highly skilled workers.

Reply | Parent | Thread

Simon Funk

(no subject)

from: simonfunk
date: Dec. 16th, 2006 12:01 am (UTC)
Link

I'm living in Mt Eden until Feb 5, so let me know if you're in Auckland for any length of time.

I would also love to hear about Rarotonga. As a Kiwi, you're allowed to stay there indefinitely if you like, no? I wonder if that applies to mere NZ residents.

"I know how I'd go about doing it and I'm reasonably confident that I could succeed."

And of course I would love to hear an overview of this...

Reply | Thread

mathemajician

(no subject)

from: mathemajician
date: Dec. 21st, 2006 11:11 pm (UTC)
Link

I'll be passing through with my girlfriend. I haven't sorted out when yet, but we might be able to meet. (I'm a bit short on time in general during my NZ visit)

No, Kiwis can't go there to live, but they can come here to live. I guess that's to stop rich kiwis simply buying the whole island and effectively killing off the local culture.

As for AI. I keep on seeing papers on neural networks and cortex that makes me think that with a few key advances, the whole field of AI could explode all of a sudden... hard to know when, but I think it should happen in the next 15 years.

Reply | Parent | Thread

(no subject)

from: anonymous
date: Dec. 24th, 2006 03:19 pm (UTC)
Link

"I know how I'd go about doing it and I'm reasonably confident that I could succeed."

is a bit stronger a claim than

"I keep on seeing papers on neural networks and cortex that makes me think that with a few key advances, the whole field of AI could explode all of a sudden... hard to know when"

Maybe if people released all the wacky ideas they've been hoarding into the public domain, the field would have exploded by now ;)

Reply | Parent | Thread

(no subject)

from: anonymous
date: Dec. 24th, 2006 09:38 pm (UTC)
Link

The two claims are about different things. One is about what I think I could do given the resources and freedom to do what I think is needed. The other is about what I think is likely to actually happen.

(comment by mathemajician... who is having trouble logging in on this crapy internet explorer and internet connection)

Reply | Parent | Thread

(no subject)

from: sapphire_schist
date: Jan. 15th, 2007 04:16 am (UTC)
Link

I just knew that Economics major would come in handy one day!!!!!!!!!

Reply | Thread

(no subject)

from: anonymous
date: Jan. 27th, 2007 06:54 pm (UTC)
Link

Hello mathemajician,

I wonder what kind of approach to AI are you considering. I am currently doing research in the same topic. I study AI applied to mathematical problems and strictly from an algebraic point of view. I have resources to live for at least five years, and for first time I am 100% on it. I know it is pretty risky to do what I am doing. I am very ambitious, but I am not naive. I know it will not be real AI any time soon. I have been thinking on it for eighteen years now and I know how hard it is. I will be happy if I get some strong theorem in the next couple of years.

I have met recently a number of people who claim they are going to have real AI in less than five years. Everybody says "less than 5 years" :-) I don´t think so.

Let me know a little bit more about your project.

Best regards
Fernando

Reply | Thread

mathemajician

(no subject)

from: mathemajician
date: Jan. 29th, 2007 12:08 pm (UTC)
Link

I actually said "proof of concept" in five years if I had the funding, not "real AI". It's not the same thing.

The approach I have planned is based on neuroscience. I think that is the most likely to lead to AGI in the near term future, i.e. the next twenty years. Indeed I'd bet money that it will. I think it's just a matter of time before the understanding of cortex and computer power get to a point where this happens.

Anyway, rather than just talking about it, I'm currently talking to some guys about starting a project to make this happen.

Reply | Parent | Thread