[MUD-Dev] New Scientist Article...

[rayzam]

----- Original Message -----
From: "Alistair Milne" <vw_krug at ntlworld.com>
To: <mud-dev at kanga.nu>
Sent: Wednesday, February 07, 2001 7:02 AM
Subject: RE: [MUD-Dev] New Scientist Article...


> Eric Rhea said:

>> http://www.newscientist.com/features/features.jsp?id=ns22751

>> I thought that this article might provide some with inspiration
>> towards whatever project they are working on. The subject of the
>> article covers "mirror" neurons, the researchers tests, and how
>> language might be involved.

> So to distil it a bit, scientists have discovered that the same
> neurons we use for doing stuff, can also change the subject from
> 'self' to 'other' and then simulate stuff instead.

Actually, it's a bit more complicated and a bit less complicated than
that :) In the exact case:

  Lift a stick from the table.   -  a neuron fires.

  Watch someone life a stick from the table with the same arm motion -
  same neuron fires.

Does this neuron code lifting a stick? Well it fires to some other
similar movements, but not all movements. Point being that half the
brain was thought to be 'motor' or related to controlling
movements. That being said, there are lots of areas outside of the
small strip known as the primary motor cortex, that fire to a variety
of movement-related activity, sometimes before, sometimes
during. Bottom line, there's a lot going on, and it's labelled vaguely
as 'movement', but we don't know how exactly we generate complicated
movements like that.

Simulate: mirror neurons are an amazingly interesting phenomenon. Most
of the neurons 'motor' neurons are dependent on own body
movements. Having it fire when the monkey sees someone make a motion,
that if he did the same motion, would also make it fire, is amazing! 
[especially if you can parse that sentence]. It's quite
surprising. There's lots of speculation over what it could do, but no
evidence. We cannot tell what the monkey is thinking, what is going
on, or even really what the neuron is doing. But that doesn't hamper
models, since the neuron shows that the brain is capable of making
this connection, that is mirroring others movements and our
own. That's the reward here.

Other interesting phenomena:

    The frontal 'motor' area that controls eye movements can also
    control your attention without moving your eyes.

    One of the motor areas for arm movements is affected by solving
    mazes.  Starting point is the center of the maze. There is an exit
    on all 4 sides.  Solve the maze to see which exit is the real
    one. If the neuron prefers arm movements to the right: if the exit
    is on the right, the neuron fires more.  If the exit is on the
    left, it fires less. And that's without moving the arm at all.

    Other arm-related neurons fire when objects are near the arm.
    Interestingly, these same neurons extend their properties along
    the length of tools that the monkey is familiar with using.

Thus, these 'motor' areas are becoming noted for being involved in a
variety of cognitive processes. Surprising in each phenomenon. But not
surprising, if the alternative is that a large portion of our brain is
used to control movements, but at the same time, many creatures with
much smaller brains can control the same or even more movements than
we can.

Mirror neurons could be used for rehearsing. They could be used for
learning. They could be related to imagination [is the monkey
imagining picking up the peanut to eat himself].  They could be part
of the theory of mind [knowing what emotions/thoughts others are
feeling based on their actions and how you'd feel doing the same
actions].

The activity could be getting ready to do the same action.
Social/group/herd animals. One senses or learns about danger, and
starts escaping. See a confederate do that, and you crank up the
neurons that would do the same thing, so you're more prepared to do
the same thing. Being more prepared neuronally means that you can
start that action quicker. And that could save you from a predator, or
other danger.

> Should we use this concept in AI?  Absolutely.  Can we do it yet?  Not
> until the field of cognitive modelling matures a bit.

I'm not sure mirror neurons are useful in AI. It could be another way
of increasing skills. Instead of skill-use, skill-watch. But that
leads to all sorts of issues. Is the smith who idles around watching
another smith practice smithing [and gaining ability out of
skill-use], going to gain smithing ability, and if so, at what rate?

I don't know how else it would be useful in game AI however, if you
could give expand on that, I'd love to hear it.

> PS: for more on cognitive modelling as applied to games, check out
>  http://www.amazon.com/exec/obidos/ASIN/1568811039/o/qid=981546964/sr=8-1/ref=aps_sr_b_1_1/107-0617214-9198949

Is the book any good? Opinions on it?

    Rayzam



_______________________________________________
MUD-Dev mailing list
MUD-Dev at kanga.nu
https://www.kanga.nu/lists/listinfo/mud-dev