Re: General summary of FAI theory

From: Byrne Hobart (
Date: Tue Nov 20 2007 - 23:27:59 MST

> It is straightforward (not easy, but straightforward) for any
> dedicated, reasonably intelligent person to become rich in modern-day
> America. All you have to do is start a company, and get millions of
> users for a product or service, no AI necessary. See the essays at
>, in particular
> and

PG usually claims that the average startup fails: that if you do
everything right, you have a good chance of ending up earning exactly
nothing (which may be worth something, especially on a résumé). In
that context, using an AI to predict the market sounds like a cost-
effective solution (and a scalable one -- I'm not sure PG's advice is
worse at turning $10,000 into $1 million than it as at turning that
$1 million into $100 million.

On the other hand, it's not like Dr. Evil won't have competition. I
can think of three firms that, collective, run about $50 billion with
over 100 PhD-educated researchers cranking out predictive software,
and even a very smart AI may not have the combined intellectual
firepower of 100 people with 130+ IQs each, and much more experience.
> No. For that to have a reasonable chance of success, you would have to
> get competent transhumanists (if not professional AI researchers)
> writing the regulations, and the bureaucrats aren't going to let that
> happen. Otherwise, you just end up having to fill out meaningless AI
> Safety Permit Application Form #581,215,102.

True. And what's especially bad is that Dr. Evil isn't going to fill
out AI Safety Permit forms, so his overhead is that much lower.

Basically, the protection you have to have is that the people pushing
hardest for AI are doing it so Dr. Evil won't be the first to get it
(although the record of well-meaning people with enormous power is a
lot shabbier than that of self-interested people with the same -- Dr.
Evil might extort money from us, but he'll probably do it as
efficiently as possible, and on the right part of the Laffer curve,
so we may end up happier after all).

This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:01:00 MDT