Re: Guessing?

From: JOG <>
Date: Mon, 14 Jul 2008 10:30:44 -0700 (PDT)
Message-ID: <>

On Jul 14, 5:45 pm, Marshall <> wrote:
> On Jul 13, 9:07 am, JOG <> wrote:
> > >> The greatest weakness in the entire debate, however,
> > >> is the capacity issue. Lack of computing capacity is
> > >> a complete explanation for what computers can't do (yet.)
> > <splutter/>
> > Ok, this one is just ridiculous. Lets take the bastion of good old
> > fashioned AI - chess. In the 90's the chess AI "deep blue" was
> > processing over 200 million board positions a second. That's right.
> > 200 millions every single second. Let's compare that to a grand
> > master, who can examine about 8. Yup, that's 199,999,992 less
> > positions per second than the AI.
> Hey! You've been complaining about the other side's simplistic
> analyses, but here you're doing exactly the same thing. Deep
> Blue included special purpose hardware for playing chess, as
> well as dozens of general purpose CPUs. And you're claiming
> it's looking at 25 million times as many positions per second.
> Yet, Deep Blue lost to Kasparov, and Deeper Blue only just
> managed to eke out a victory. So, the 25 million number is
> crap, isn't it?

C'monnnn, its incredible. Examining 8 positions per second vs 200 million. Anyway you put it that's astounding. In addition Deep Blue had a move memory that could instantly examine 700,000 grandmaster games as well. Phew-ee. (CYC itself ended up with over 1.4 million rules of logic and could process them at blinding speeds, but could do jack all. I don't know about you but that's more rules than I can think of).

Waving our hands about and saying "if only we just had more processors" or "if only we could just add /another/ hundred thousand rules", is just wishful thinking (bordering on religious faith). The problem is not capacity, it is approach. I agree that there is no reason to think AI is not achievable, but theory suggests formalized descriptions of problem domains aren't enough. Experiments seem to be giving us supporting evidence.

> It appears the human and the machine are
> roughly at parity, in terms of chess playing ability. And Kasparov
> is a general-purpose machine, capable of doing things that
> Deep Blue can't, like ride a bike.
> One lone human has *way* more processing power than today's
> computers. This is a sufficient explanation of their difference
> in cognitive ability.

Someone claiming progress each time they add a rule or processor cycle to a system, is no different to them climbing a branch higher in a tree and claiming progress in a quest for the bloody moon.

Tell you what you take the tree, the rest of us will try and work out some other transportation ;)

> Marshall
Received on Mon Jul 14 2008 - 19:30:44 CEST

Original text of this message