artificial intelligence

(my replies interspersed throughout below)

--- "Omar E. Vega" wrote:

> > There has been some advances in this field..
> Standford's DARPA
> >Grand
> > Challenge entries the last few years have been
> pretty impressive..
> > Honda has made some progress too. as well as
> others.. Good image
> > recognition is a huge challenge, but you have to
> figure a five year
> > old child has been working on his/her programing
> for 5 years
> > straight.. And that's after evolution has already
> done most of the
> > work with the parents supplying all the proper
> equipment..
> >
> Really? Do you mean, those metalic puppets shown on
> Japanese
> universities are an advance?

I'd say that bipedal robots that can run (such as
Asimo) are a huge advance over the state of the art of
even ten years ago (Genghis and Atilla). Even some of
the smaller type of robots, the robosapiens-style
ones, are doing some amazing things in comparison to
the robots of five or ten years ago. And as far as
them being "puppets" goes, what is it that you want?
Should they simply be fully intelligent at the level
of an adult human being right out of the box?

I think that if you were to show the robots of today
to someone who was at the top of the AI field 25 years
ago, they would be amazed at how far we have come, and
would say that we have definitely achieved a form of
AI already today.

>
> Remember with robot vision we are not talking about
> the complex,
> self organizing task of the brain at all. We are
> talking about BASIC
> TASKS. Very BASIC tasks. Thinks that the brain do by
> instinc,
> without even thinking about it. And certainly
> animals like dogs and
> cats do as well.
>
> Do you know there still no segmentation algorith
> that is failure-
> proof? For example.

There is no such thing as failure-proof. This is what
I meant by "moving the bar". In this case, the bar
has been moved beyond that of which even human beings
are capable. If you have ever been fooled by an
optical illusion, then the circuitry/algorithms in
your eyes and brain are not failure proof. Is the bar
for AI to be set so high that not even human beings
could be considered intelligent?

>
> That's what I mean when I say scientists
> (particularly A.I.
> researchers) are very good for saling "dreams",
> instead of finding
> solutions.
>
> Modelling the mind (that's what we are talking
> about) is a task that
> will take centuries, to achieve, perhaps thousand of
> years. Before
> that's achieved I am sure predicting "A.I. for the
> next five years"
> is not only the saying of charlatans but is cheating
> people.

AI is already here. Look around you. It is so
pervasive that modern life would be impossible without
AI. Sure, a lot of it is in specialized systems -
like the railroads of Japan, or in petroleum
refineries, or the Google search engine - and it isn't
the cartoonish version envisioned in the science
fiction of 50 years ago. That doesn't make it any
less intelligent.

>
> Singularity will have to wait a little.
>

I don't think the singularity will be a single event;
instead it is a process that will occur gradually over
several years along many different fronts, and it will
happen in fits and starts, and that it is happening
right now.

Look at it this way: when was the Renaissance? On
what date did it start, and on what date did it end?
If you were alive early in the Renaissance, would you
even realize that it was happening, even if you were
one of the principal figures of that age?

It was only in hindsight that we were able to look
back on that period of time and give it the
Renaissance label. I think it will be the same with
the Singularity, and that in 20 years or so we will be
able to look back and say, "hmm it looks like we
passed the Singularity back in about 2004 or so..."

Ed
(former researcher in neural networks, genetic
algorithms, and fuzzy logic, no kidding ;) )