[ExI] Oxford scientists edge toward quantum PC with 10b qubits.

[Kelly Anderson]

On Fri, Jan 28, 2011 at 9:01 AM, Richard Loosemore <rpwl at lightlink.com> wrote:
> John Clark wrote:
>>
>> http://www.youtube.com/watch?v=WFR3lOm_xhE
> Yes, but do you have any idea how trivial this is?

Trivial!?! This is the final result of decades of research in both
software and hardware. Hundreds of thousands of man hours have gone
into the projects that directly led to this development. Trivial! You
have to be kidding. The subtle language cues that are used on Jeopardy
are not easy to pick up on. This is a really major advance in AI. I
personally consider this to be a far more impressive achievement than
Deep Blue learning to play chess.

> The IBM computer playing Jeopardy is just a glorified version of Winograd's
> SHRDLU.  With enough information it can home in on answers to simple
> questions.  Doing that kind of stuff is like winning a barellized
> fish-shooting context:  if you have a big enough encyclopaedia in there, and
> you do a fast enough search, you get near to the relevant facts. But that is
> not the same as structured intelligence.

Whatever you want to call the technological form, or despite the
geneaological roots of this particular software/hardware, it is not
trivial, and it is not shooting fish in a barrel. Having the facts,
yes that is easy enough. Gathering the facts from natural language
sources, parsing natural language questions (answers in this case) and
answering them (questioning them in this case) is pretty dang cool. I
would not call it trivial.

> As I write these words I am sitting here getting ready to teach some
> students enough physics and vector calculus that they can understand Gauss's
> theorem, Maxwell's equations, the subtleties of EM induction ... and these
> kids will (if we're lucky) be able to understand all that in a couple of
> months' time.
>
> But if I tried to have a conversation with that IBM Jeopardy computer about
> these things, would it be able to start understanding, if I took it real
> slow?  No, not at all.  If you know something about the techniques and the
> tricks that the IDIOT BLUE team are using to get their baby to do that
> stuff, you will know that this is not a step on the road, it is a dead end.

Here you may be right. Strong AI will probably not be based on this
technology. However, I would certainly rather speak to Watson than the
typical idiot I get on the phone on the typical technical support
call. I think that's what IBM is really about here. You don't need
Strong AI to answer domain specific questions like that, and this
software apparently can do that. Granted, it's a lot of hardware
today, but Moore's law will take care of that shortly.

> What is sad about all this is that AI has been through so many of these
> cycles.  Thinking that dictionary lookup plus a few extras is all you need
> for intelligence.  This is true.  It is just that the "few extras" are
> 99.999% of the problem.

Richard, do you think computers will achieve Strong AI eventually?

I do. I think it will likely come from reverse engineering brains, or
perhaps creating bio computers like those fun little robots that run
around on rat brains (see Youtube).

-Kelly