RE: Einstein, Edison, & IQ

From: sl4ghu@mattbamberger.com
Date: Thu Nov 17 2005 - 23:29:39 MST


Michael Anissimov wrote:
> At least one person I know of,
> Matt Bamberger, started independent work on AGI this year.

Minor correction-- I'm not really working on AGI in any serious way yet.
Right now, I'm finishing up some prior obligations, and trying to gain
enough hands-on experience with AI to make a rational decision about what to
do next. Once those things are done, I might end up starting a new project,
although my first choice would be to help out on an existing effort.

I agree with the general sentiment that finding good people is super-hard.
In my present job, I spend a fair amount of time hiring programmers. I'm
currently trying to hire a couple of people for one of the best-known and
most successful game companies in the world, offering highly competitive
salary and benefits (we'll go head to head with MS or Google). Our only
requirement is that we want world-class programmers-- we'll look at top
people with any experience level, from college students to 30 year veterans.
Even so, we're having an incredibly hard time finding people. That's pretty
much universal in the software industry.

For AGI, the problem is much harder. The quality bar is similar (maybe much
higher, if you believe SIAI), the pay is likely to be much lower, you need
to be sl4 but not a nutjob, you have to have some knowledge of AI, and you
have to be willing to work for a company that nobody's ever heard of, doing
work that most people think is crazy, with dubious prospects for ever
getting rich.

My gut estimate (I believe SIAI's estimate is higher) is that the absolute
minimum team size for a realistic AGI effort is 6 full time senior-level
programmers. I believe that's more than the combined current teams at SIAI,
Novamente, Numenta, and A2I2.

I suspect that solving this problem will necessitate aggressive outreach and
recruiting within the broader tech community, not just singularitarians.

        -mattb



This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:53 MDT