Re: rambling question on AI

From: Eliezer S. Yudkowsky (
Date: Sat Jan 05 2002 - 08:12:22 MST

Reason wrote:
> I'm not convinced that any project is so intrinsically hard that you can't
> find comparatively rapid substeps to producing something profitable (or at
> least salable) on the way. If you can get non-stick frying pans out of the
> space program, then I'm sure there's something you can get out of AI
> development.

The key word is "comparatively" rapid. I can think of comparatively rapid
routes to spinoffs; I cannot think of routes that are absolutely rapid,
i.e. rapid by comparison with an average programming project.

There's plenty that can be got out of real AI development, so long as it
is real AI and not something else.

> How long did Cyg take to get to a point of being saleable? You could, for
> the sake of argument, view that as an attempt at a first step (i.e.
> producing the database) in a much longer ongoing project.

Cyg, or Cyc? I don't recall offhand an AI project named Cyg. And Cyc is
not exactly the best example you could pick to make your case.

> We should all be trying to get Flare slashdotted, by the way :) Some more
> attention couldn't hurt.

I think the current thought was that we wanted a prototype or a whitepaper
before we tried to get Flare slashdotted. Right now we need coding
volunteers, but we definitely don't need to use up our shot at Slashdot
while the project is still revving up - the attention would probably do
more harm than good.

> > Furthermore, no matter how good an idea I have for a for-profit
> > corporation, it still isn't going to have better than a fifty percent
> > chance of success, because the ordinary failure rate for startups is
> > ninety-five percent.
> I'd say it has a 95% chance of failure. If there's one thing that my time in
> the Bay Area has taught me, it's that how good you are, how good your
> products are, whether or not you have the Best Idea; none of these matter a
> damn. The success or failure of your company is largely out of your hands
> and lies in the variables over which you have no control.

Yeah, pretty much. I don't think that *all* the variables are out of
control, just a heck of a lot of them, so I do think it's possible to
increase the chances somewhat. But for the record, I don't have any
specific idea that I would regard as giving a company a fifty-percent
chance of success; I was trying to quantify a theoretical maximum.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 13:37:32 MST