Re: Sony to quit manufacturing AIBO, QRIO


Wolf Kirchmeir wrote:
> Traveler wrote:
> [...] The public wants robots that can do chores around the
> > house, chores that require sophisticated intelligence. We're not there
> > yet. But it's coming. The good thing about all this is that we now
> > have good robot hardware technology.
> [...]
>
> The successful replacement/augmentation of some human skills and not
> others shows that what's easy is not necessarily simple, and what's hard
> is not necessarily complicated. An obvious principle, once you recognise
> it, but widely ignored. Logic and math are hard for us, but simple.
> Cleaning the house is easy for us (hence its tedium), but complicated.
>


Minsky and Simon and others have been saying this for decades. Actions
that require "common sense" are tons more difficult, since they require
so
much background knowledge, and ability to process many forms of sensory
data. Narrow AI is brittle. It takes a general approach to AI for
success
in general situations.



>
> I find that many comments about "intellignece" tacitly assume that tasks
> that are difficult for humans must require more intelligence than easy
> ones, despite much evidence that it's not so. I suppose that's because
> so many workers in the field have worked very hard to become good
> "symbol processors", and get paid a lot more than the guy or gal that
> empties their trash cans. That's why I from time to time reiterate my
> suggestion that AI should try emulating something like an ant. It's not
> at all obvious how to make a robot as smart as an ant, despite the fact
> that we know a good deal about ant behaviour. It's easy enough to make a
> toy that walks like an insect; but that's nowhere near a robot ant.
>
>


This is a bare start, still not much of a solution, for the reason that
the ant
has an extremely limited sensory-action repertoire. It has nowheres
near
the "common sensical" power to survive in an environment a child can
operate in. It's strong and cute, but it's lifestyle is pretty obring.
The ant
approach is as brittle as GOFAI solutions in general situations. EG,
subsumption bots are robust in specific environments, but cannot
do much overall. They're the *bottom* level, and many more levels of
processing intelligence are required on top of that.


>
> As for Sony's robot dog: It's slightly more interesting than those toys
> that walk like an insect. For about 5 minutes.
>
>


This is the basic problem. And the iCybie's were even worse, since all
they could do were a few tricks, and they couldn't even do a decent
walk.

.

Relevant Pages

  • Re: Sony to quit manufacturing AIBO, QRIO
    ... chores that require sophisticated intelligence. ... >> have good robot hardware technology. ... >Cleaning the house is easy for us, ... >suggestion that AI should try emulating something like an ant. ...
    (comp.ai.philosophy)
  • Re: Where is behavior AI now?
    ... purpose real time reinforcement learning. ... I'd love to see a robot with the capicty of, say, an ant. ... Collecting trash might be another interesting application for smarter ...
    (comp.robotics.misc)
  • Re: Sony to quit manufacturing AIBO, QRIO
    ... >>> out that knowing which buttons to push on the ... > students were shown why it worked. ... >>> a robot ant, then a few thousand of those robot ...
    (comp.ai.philosophy)
  • Re: Sony to quit manufacturing AIBO, QRIO
    ... >> out that knowing which buttons to push on the ... students were shown why it worked. ... >> a robot ant, then a few thousand of those robot ...
    (comp.ai.philosophy)
  • Re: Fantastic Four
    ... > As in Harrison Ford in big neon city with hot robot girl fighting big robot ... > guy in weird house full of walking dolls? ... I can usually appreciate a good film even when I don't like it, ...
    (uk.games.video.misc)