Re: Professional intuitions

Eliezer S. Yudkowsky (
Sat, 26 Sep 1998 20:24:09 -0500

Damien R. Sullivan wrote:
> Tangent: Robin studied physics from 1977 to 1984, and worked in AI from 1984
> to 1993. *Then* he switched to economics. This is from his web site. I
> don't know where he feels he has more powerful intuitions, but I wouldn't
> assume they were in economics. Or that your intuitions are better than his,
> at least based on experience.
> [...]
> Specialty? What's Robin's specialty? He's currently focused on econ, but he
> spent half your life in AI.

Let's not get into this. I made a basic point about the different perspectives of different people in different fields; perhaps I should have made a point about the different perspectives of different people, period, but I thought there was a correlation involved. I didn't imply that I was any better at it than Hanson. I said that there were meta-reasons why communication would be difficult. I didn't challenge anyone's qualifications or imply an asymmetry; quite the opposite.

Arguing about who has more experience is an argument I can't win, almost by definition. If you look at the people who understood enough to change the world, the amount of experience over and above a few years of hard work is irrelevant. We all know who and what I am, or think I am, because I said it the last time I was here; I don't intend to fight fire with fire, or try to shift the battle to my home ground instead of yours. Comparing the qualifications of two very bright people who each have years of experience is nothing but an ad hominem challenge, and this discussion is doing too well to deserve that.

> > The Big Bang was instantaneous. Supernovas are very bright. State-vector
> > reduction is sudden and discontinuous. A computer crashing loses all the
> > memory at once. The thing is, human life can't survive in any of these areas,
> > nor in a Singularity, so our intuitions don't deal with them. The Universe is
> I'm not sure what analogy you're trying to make here. The one I'm
> constructing out of these instances is "Yes, the universe is full of many
> sudden phenomena. Big, simple, destructive phenomena. But the destruction of
> the Singularity is a side-effect; what it *is* is a sudden increase in
> complexity. A creative discontinuity, not a destructive one."
> And a sudden jump in complexity is implausible. The closest analogies which
> come to mind are crystallization and the probable spread of bacteria through
> the early oceans. I'm not sure the former is valid; as for the latter, we
> don't know for certain what happened, and there's a unique qualitative change
> there: the rise of self-replicating entities on an otherwise dead planet.

One man's sudden jump is another man's slow rise; more and more I feel we're debating timescales, rather than fundamentals. From a sufficiently slow perspective, this planet went from bare rock to intelligence in a mere fraction of a moment. We all know that a 333 Mhz PowerPC G3 chip operates a million times faster than the neurons in the brain, on which scale a year takes only thirty seconds; and likewise that chip has a billionth of the capacity of the brain, on which scale thirty seconds takes a millennium. It is rather unlikely that a slow, steady advance from the _AI's_ perspective will appear to us as anything but a flat line or a vertical rise.

The Universe is full of stable things, saith Richard Dawkins. There are exactly two stable states in the Universe: Dead matter, and superintelligence. Unintelligent life persists for a few billion years at the most, but dies with its star. Mortal-level intelligence destroys itself or falls into a Singularity. Dead matter generally remains dead, so it is stable; superintelligence can preserve itself, and whatever its progresses and internal changes, it remains "superintelligence" from our perspective.

So from the perspective of a Universe-watcher, the Universe consists largely of clumps of matter. Occasionally these clumps of matter flash into Singularities.