Re: Protean Self-Transformation

David McFadzean (david@kumo.com)
Mon, 31 Mar 1997 10:05:52 -0700


At 01:13 AM 31/03/97 -0600, Gregory Houston wrote:
>David McFadzean wrote:
>
>> What if the images that the computer analyses were generated with
>> a ray-tracing program? No optical hardware needed.
>
>That would be much more like imagination than seeing. I do not require
>eyes to see things in my head that I make up, but if I want to see
>things from the real world then I need something that will recieve
>information from the real world.

If it was like imagination then the vision program would be able to
change what it was trying to see. Since the images come from outside
the program (perhaps they were generated years ago by another computer
and are merely stored on disk), they are an objective feature of the
program's world. It is not at all like imagination.

>> Randall Beer and his students successfully taught (programmed/trained/
>> evolved) a simulated cockroach to walk though it had no real (hardware)
>> legs. When the neural nets were later downloaded into a real robot it
>> could instantly walk in the real world.
>
>It was trained to walk, but it could not walk until it had the robot
>embodiment. It could think about walking all day long, but until the
>instructions were downloaded into the hardware (the robot) there was no
>walking going on.

If it looks like it is walking (it moves around its world) and thinks
it is walking, then what is the difference? I suppose you could define
"walking" narrowly enough to exclude what the simulated cockroach is
doing, but that is like saying computers don't do math until the results
are printed on paper.

--
David McFadzean                 david@kumo.com
Memetic Engineer                http://www.kumo.com/~david/
Kumo Software Corp.             http://www.kumo.com