SIAI will self-destruct

From: David S. Hansen (haploid@haploid.com)
Date: Fri Jun 04 2004 - 14:56:09 MDT


With the recent SIAI-bashing on sl4, I might as well jump into the fray.

My prediction:

SIAI will self-destruct. It will fail to attract substantial funding.
It will lose in the race to the Singularity. SIAI will still be working
on movie website tie-ins, will still be marginally funded by college
students and the occasional eccentric wealthy individual, and Eliezer
will still be hammering out his friendliness theory with absolutely no
working code in the year 2025 - about the time a generic bayesian
knowledge network bootstraps itself into superintelligence.

1. Ben's points about the arrogance of Eliezer and Michael are *very
much* relevant to the ability of SIAI to meet its goals. There are a
couple orthogonal reasons for this:

1a. Arrogance tends to close one's mind to the consideration of outside
ideas. An idea monoculture can and usually does severely limit an
organization's ability to succeed. Ackoff and other organizational
systems theorists have demonstrated this extensively.

1b. Arrogance, even if it is "correctly-placed" and "deserved", often
limits an organization's fundraising capacity. Acting like a complete
ass is one very effective way of dissuading people from allocating
resources in the direction of said ass. This effect is *significantly*
amplified if the principals in question have no external claim to the
basis of their arrogance( e.g. high honors, doctorate degrees, industry
success, etc ). Serious investors do not buy into an idea on faith alone.

1c. SIAI's apparent mistrust of "outsiders", as demonstrated in the
oft-repeated claims that LOGI is obsolete and there is something much
better but it can't be published because evil people will implement it
before we get friendliness worked out. This position has the direct
effect of discouraging talented potential seed AI programmers from
becoming interested in the project. It also has the direct effect of
discouraging outside investment due to disclosure and control concerns.
  It also has the direct effect( and I believe Ben has already noted
this )of discouraging other AGI researchers from contributing ideas
toward the development of SIAI's theory and project, e.g. "why bother if
what they've published is useless anyway?"

2. Not only is there a complete lack of theoretical cohesiveness
amongst AGI projects( e.g. Eliezer and Ben ), but there is a complete
lack of theoretical, structural, and motivational cohesiveness amongst
the principals of SIAI itself. I don't think a day goes by( recently
)during which Eliezer doesn't correct Tyler or Anissimov on at least one
very critical point. Acknowledging the fact that *correctness* is
important to maintain, I would argue that when attempting to gain public
interest or outside funding, this is a critical flaw. Seasoned public
relations professionals know that if one doesn't know the answer, one
doesn't make it up. When attempting to secure funding, one doesn't
present to the public a group of highly disconnected individuals.

3. Awareness Efforts. Admittedly, I was a little excited about the
flurry of activity that Tyler initiated, as I got the impression that
SIAI was finally getting serious about promoting itself. This is what
caused me to begin donating and helping in the first place. After a
while, however, it became clear to me that it was more of the same.
Undertaking herculean efforts in order to drive a few more hits to a
website. Bringing on a 20-year-old with no significant network to go
fundraising in silicon valley. Wow. What SIAI seems to not understand
is that when one undertakes a serious round of fundraising, one must
approach potential serious investors with either credibility or a
product - and for reasons noted above, this is something SIAI does not have.

-- 
o David S. Hansen
o haploid@haploid.com


This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:47 MDT