[ExI] AI Motivation revisited

[Eugen Leitl]

On Fri, Jul 01, 2011 at 07:11:33PM +0200, Stefano Vaj wrote:

> As long as you go on, replacing along the way dead humans (and perhaps dead
> planets, since the time I suspect it would take) what's the problem?

The problem is that it doesn't happen, in practice. The longest
supercomputer batch jobs I know of are all 2-3 years (and have
to be re-run in order to validate them, despite all the error
correction tens of thousands of nodes running that long will
produce uncatched errors).
 
A typical run takes days to weeks. Most supercomputer
installations have a lifetime of much less than a decade
without upgrades. Business upgrade cycles are 2-3 years,
as older hardware is not worth the juice to keep running.
 
> Floppy is 360 kBytes, and for pracical reasons you can't use more
> > than 10 k of these, so it's about 4 GByte of state.
> > A 360 kBytes floppy writes at about 32 kBytes/s, so you
> > need over 100 days just to write these once. Add handling,
> > wear, diagnostics and error correction and you can
> > easily multiply that by 2 or more.
> >
> 
> Why on the earth can't I use more that 10k floppies? And what is 100 days if

Because you can only write that floppy set about ten times in a decade.
That assumes you're streaming through, with random access across the
entire data set you can be lucky to write once in a century.
How does less than a second versus a century sound like? 
And how does a nanosecond (dedicated hardware) versus a century 
sound like? Ok, that was computers. With people, multiply this by >10^7.
That's a gigayear or ten. And for some strange reason people (ok, for small
values of people: philosophers) are taking *Searle* seriously.
Lolwhut?

> I have all the time of the world? My point is not that the original PC would

Nobody has all the time in the world. As I said, human affairs are
ephemeral. I can guarantee you one thing: the only way the 10 kiloyear
clock is going to manage anywhere near that runtime is when we all
get raptured. 

> be very performing.

It would not just be not very performing. It would never finish.
 
> My point is that we would still recognise as "sentient" a man slowed down by
> a factor of 10x, 1000x, 10^10x, etc., so that as a pure Gendankenexperiment

Factor 10, no problem. Factor 10^3, some problem. Anything beyond that,
and you're looking at a multigeneration project. Oh, a bright a shiny
object... say, what's that statue doing there? And just who vandalized it
so brutally?

> my example is valid.

My point was that in theory, there's no difference between theory
and practice. In practice, there is.
 
> 
> > > OTOH, I am by no means certain that organic brains are so poorly
> > optimised
> > > to run "AGI" programs in comparison with other conceivable, eg, silicon,
> > > supports.
> >
> > It's the opposite. When Sun (vanquished by Orkcackle) said
> > that the network is the computer they spoke truer than they
> > knew. Silicon has lousy fanout, and modern systems are
> > extremely poor at parallelism. Which is why you need huge
> > clusters with millions of cores to do something quite
> > trivial.
> >
> 
> Yes, we fully agree on that (even though the typical counterargument is that
> organic brains have been molded by evolution and natural selection, which
> could have constrained us in some less-than-ideal paths in the space of all

Of course it's less than ideal, taking a theoretical limit. But in
practice the hardware is both extremely powerful and extremely efficient,
given the energy constraints.

> those theoretically possible, and which could be adopted by design).
> 
> Conversely, organic brains are not so brilliant with basic arithmetics.

It does plenty of basic arithmetics very quickly if you consider analog
computation. If you consider the realtime inverse kinematics you're
performing you'll be beating a supercomputer. 
You, as a top layer of the process don't realize this, and of course
we didn't have time to evolve to deal with symbolic arithmetics on
the top level.

-- 
Eugen* Leitl <a href="http://leitl.org">leitl</a> http://leitl.org
______________________________________________________________
ICBM: 48.07100, 11.36820 http://www.ativel.com http://postbiota.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE