Chris Pfeiler
2009-02-06 17:28:11 UTC
Now here is something interesting (I hope) for the small community.
I´m in email contact with Sandra Linkletter and I asked her of course
about her final post here and her reasoning in it that the norns were
robbed of their possibility to become true AL. I have her permission
to post parts of her mail here and I´m very grateful for that.
BTW, regarding my plan to bring back some ontopic-business to agc,
she has this to say: "Yes, why not drop a pebble in the still pond,
and watch the ripples spread. *laughs*"
So let them spread. Here are some detailled thoughts on C1, C2 and
AL in general from slink:
---------
Creatures 1 was on the right track for what Steve Grand envisioned,
which was an virtual life-form which could evolve into forms most
fit for their environment. There were two flaws in the model, in C1.
The first was that the brain was not expandable, and to a minor degree
behavior was hardwired to certain responses. One can see that when a
norn is stuck in repetitive behavior, and it's motor control messages
are sending "left" over and over again. I think this had something to
do with the interface between the senses and the brain.
The second was the lack of mass-energy constraints. Mutations which
created matter from nothing could occur. When this happened, the
norn possessing the mutation had a great advantage over otherwise much
fitter norns. As an overwhelmingly positive mutation, it caused the
perpetuation of strains of norns with painful and even lethal medical
conditions. Thinking back, I feel great pleasure over the studies
that I and others did of the C1 norns.
I am not certain that anyone but Steve Grand understood the second
problem, and he was heavily criticized in the academic AL circles
for the lack of mass-energy constraints. They phrased it in jargon
of their trade, but basically what they said was that if you leave
a hole in a container, water will always find it. That is not helpful
when one is trying to study the behavior of water in a container.
C2 promised to remedy the first problem. However, the promise of C2
was compromised by the realities of the marketplace. First of all,
they had to design a new system from scratch because of their split
from the portion of the company who owned the rights to the original
program. I believe that was Millennium? Then, well into the project,
they decided to completely rewrite the graphics engine because it was
too slow as it was. This set everything else back at a time when the
game only had a few months to release.
They had a genuine neural network expert on their staff, but I believe
that nothing much was actually ever done with what appeared on paper
to be a complex and interesting model for the AI. I believe this was
due in part to time constraints, and also because their game engine
could not support all the eye candy and also do a proper job on the AI.
I remember getting complaints because my pre-birth instinct genes took
more than 30 seconds to run. They said Creatures was supposed to be an
interactive game, and players would be bored if the eggs did not hatch
immediately when placed in the incubator.
I probably made the biochemistry overcomplicated for a game program. I
confess to that, but I feel it was a good piece of work regardless.
The appearance genes were as many in number as the biochemistry genes,
which reflects the priorities of CyberLife. When the game was released,
it was quickly apparent to me that they had coded in behavior that
should have been learned. That is, they had turned AL into a virtual
pet. Furthermore, it was a virtual pet that died quickly in many ways,
because it was supposed to be learning how to live and not following
some predetermined pattern.
One wry comment that I can make is that the behavior of a species of
animal should not be coded by a team of single young men, working
over one weekend to decide the proper priorities of needs, who have
no personal experience with keeping even so much as a pet dog.
It was possible to change the behavior of a C1 norn by editing the
brain lobes, but not as much as it seemed it ought to be. It is
possible that the degree to which the C1 norns were hard-coded was
much higher than I realized. However, I do believe firmly that
Steve Grand meant for them to be more than a game. I believe that
the neural network model in the white paper for C2 was a valid model.
What Toby Simpson did with it is obvious. It's very possible that
Cyberlife always meant to milk Steve Grand's professional reputation,
and never make a true AL. They never even granted that I had one to
milk, which could be a source of bitterness to me if I had not moved
on in life.
Whether or not the C2 norns could ever have become conscious, had they
been made properly, I don't know. Some people still believe that
animals aren't conscious, so it is apparent that the very word has
problems of definition.
-----
Thanks again to slink (if she is reading this post) for the permission
to post it here.
Chris
I´m in email contact with Sandra Linkletter and I asked her of course
about her final post here and her reasoning in it that the norns were
robbed of their possibility to become true AL. I have her permission
to post parts of her mail here and I´m very grateful for that.
BTW, regarding my plan to bring back some ontopic-business to agc,
she has this to say: "Yes, why not drop a pebble in the still pond,
and watch the ripples spread. *laughs*"
So let them spread. Here are some detailled thoughts on C1, C2 and
AL in general from slink:
---------
Creatures 1 was on the right track for what Steve Grand envisioned,
which was an virtual life-form which could evolve into forms most
fit for their environment. There were two flaws in the model, in C1.
The first was that the brain was not expandable, and to a minor degree
behavior was hardwired to certain responses. One can see that when a
norn is stuck in repetitive behavior, and it's motor control messages
are sending "left" over and over again. I think this had something to
do with the interface between the senses and the brain.
The second was the lack of mass-energy constraints. Mutations which
created matter from nothing could occur. When this happened, the
norn possessing the mutation had a great advantage over otherwise much
fitter norns. As an overwhelmingly positive mutation, it caused the
perpetuation of strains of norns with painful and even lethal medical
conditions. Thinking back, I feel great pleasure over the studies
that I and others did of the C1 norns.
I am not certain that anyone but Steve Grand understood the second
problem, and he was heavily criticized in the academic AL circles
for the lack of mass-energy constraints. They phrased it in jargon
of their trade, but basically what they said was that if you leave
a hole in a container, water will always find it. That is not helpful
when one is trying to study the behavior of water in a container.
C2 promised to remedy the first problem. However, the promise of C2
was compromised by the realities of the marketplace. First of all,
they had to design a new system from scratch because of their split
from the portion of the company who owned the rights to the original
program. I believe that was Millennium? Then, well into the project,
they decided to completely rewrite the graphics engine because it was
too slow as it was. This set everything else back at a time when the
game only had a few months to release.
They had a genuine neural network expert on their staff, but I believe
that nothing much was actually ever done with what appeared on paper
to be a complex and interesting model for the AI. I believe this was
due in part to time constraints, and also because their game engine
could not support all the eye candy and also do a proper job on the AI.
I remember getting complaints because my pre-birth instinct genes took
more than 30 seconds to run. They said Creatures was supposed to be an
interactive game, and players would be bored if the eggs did not hatch
immediately when placed in the incubator.
I probably made the biochemistry overcomplicated for a game program. I
confess to that, but I feel it was a good piece of work regardless.
The appearance genes were as many in number as the biochemistry genes,
which reflects the priorities of CyberLife. When the game was released,
it was quickly apparent to me that they had coded in behavior that
should have been learned. That is, they had turned AL into a virtual
pet. Furthermore, it was a virtual pet that died quickly in many ways,
because it was supposed to be learning how to live and not following
some predetermined pattern.
One wry comment that I can make is that the behavior of a species of
animal should not be coded by a team of single young men, working
over one weekend to decide the proper priorities of needs, who have
no personal experience with keeping even so much as a pet dog.
It was possible to change the behavior of a C1 norn by editing the
brain lobes, but not as much as it seemed it ought to be. It is
possible that the degree to which the C1 norns were hard-coded was
much higher than I realized. However, I do believe firmly that
Steve Grand meant for them to be more than a game. I believe that
the neural network model in the white paper for C2 was a valid model.
What Toby Simpson did with it is obvious. It's very possible that
Cyberlife always meant to milk Steve Grand's professional reputation,
and never make a true AL. They never even granted that I had one to
milk, which could be a source of bitterness to me if I had not moved
on in life.
Whether or not the C2 norns could ever have become conscious, had they
been made properly, I don't know. Some people still believe that
animals aren't conscious, so it is apparent that the very word has
problems of definition.
-----
Thanks again to slink (if she is reading this post) for the permission
to post it here.
Chris