Perhaps my point got missed?

I totally agree that with an AI Addon Card, that "better" AI could probably be built
for computer games.

But, is there sufficient market for buying such an additional card to be found
in the majority of computer game players? When this discussion was held years ago,
it was thought not. Has that market situation changed? If so, what evidence can
anyone offer to suggest that the typical computer game player would shell out the
$$$$ to get an AI Addon Card.

The simple economics are that if there is an insufficient market for the card, then it
is unlikely that someone would invest in developing one for our AI usage.

Eric

I know nothing about proper AI, but I believe that more "lifelike" results can be obtained with current hardware. My opinion is that a lot of contemporary developers use their available cycles and processing power rather carelessly. Black & White (for all it alleged AI shortcomings, it still is a vast improvement over most of the field) runs on the same hardware as a Quake III. Demis Hassabis (I hope that''s the man''s name) has discussed a new method he''s created for large scale AIs.

Restricting problems to a specific domain and focusing on results rather than means - meaning employing heuristics or "sythnetic" (aiming for similar results) rather than artificial (aiming for similar process) intelligence - may yield the greatest gains in the field of computer gaming. Additional hardware would only (currently) be applicable in industrial AI, where clock cycles and processing power are even more plentiful...

JMHO.

quote:Original post by Oluseyi
I know nothing about proper AI, but I believe that more "lifelike" results can be obtained with current hardware. My opinion is that a lot of contemporary developers use their available cycles and processing power rather carelessly. Black & White (for all it alleged AI shortcomings, it still is a vast improvement over most of the field) runs on the same hardware as a Quake III. Demis Hassabis (I hope that''s the man''s name) has discussed a new method he''s created for large scale AIs.

Restricting problems to a specific domain and focusing on results rather than means - meaning employing heuristics or "sythnetic" (aiming for similar results) rather than artificial (aiming for similar process) intelligence - may yield the greatest gains in the field of computer gaming. Additional hardware would only (currently) be applicable in industrial AI, where clock cycles and processing power are even more plentiful...

JMHO.

I mostly agree. One of the promises of using the AI Addon Card, would be to provide
processing power for the exclusive use of AI, so that processing cycles would not be
"carelessly" consumed by graphics and user interface. However, I disagree that some
additional hardware would only be applicable in industrial AI. Computer gaming AI was
suffering from a dirth of resources available, and even now with more publisher awareness
(I can''t call it emphasis, because its not) of AI, the resources being made available are
still less than graphics and UI. The new generation of graphics cards (GForce 3, etc.)
are even helping by taking over even more of the graphics workload than ever before.

Also, after visiting with Demis at GDC 2001 and attending his lecture, I am looking forward
to his game Revolution and the AI it is hoped it will demonstrate. Demis has some
ambitious plans and I hope they succeed.

Eric