Re: NOW: What needs doing [Yudkowsky]
Tue, 23 Mar 1999 21:11:14 -0800

Eliezer S. Yudkowsky, <>, writes:
> Concrete: The primary thing that needs doing, ASAPractical, is a
> foundation intended to turn the design in "Coding a Transhuman AI" into
> code. A few $M in seed money could create a foundation with a fairly
> respectable front, that could hire (1) me (and/or a few other
> competents) to finish the design and start turning it into something
> codeable, or more likely spend the next two years creating
> infrastructure; (2) someone to go around asking the major Silicon
> Tycoons for funding; (3) an OpenSource website to recruit the
> Manhattan-Project-like number of programmers needed.

In order to achieve this goal, a possible step forward would be to demonstrate your competence at software design. If you are capable of solving perhaps the hardest problem imaginable, the design of a software system capable of reaching super-human intelligence, then you might be capable of solving simpler design problems, as well.

It would be especially useful if you were able to design AI software which was able to demonstrate some practical capability. Look at people like Rod Brooks and Doug Lenat. Both started with small successes and were able to use their "reputation capital" as a basis for much larger projects. Once you have a track record as a successful designer, you will find it easier to attract funding for your long term project.

Even the most visionary and gifted architect will not be allowed to build a massive skyscraper as his first project. You have to start small and work your way up.