The Big Project is to try and create true AI or consciousness
on a computer. There are three possibilities:
- True AI can be implemented on a Von-Neumann computer;
- True AI can only be implemented on a computer that incorporates true
(quantum-effect)-indeterminateness;
- True AI cannot be implemented on a computer at all.
alifegames.com takes no position on which alternative will turn
out to be correct. But one thing is certain:
If we find out which, we will know a lot more about what it means to be
intelligent and conscious. We will have made a big step towards answering
one of the most basic philosophical questions: Who are we?
There are three possible pathways towards creation of true AI or consciousness
on a computer (or network of computers):
- We can try and build a thinking, conscious computer based on a theoretical
understanding of intelligence and consciousness;
- We can try to scan the brain molecule for molecule, and recreate it in a computer model;
- We can try to evolve consciousness and true AI by having
"brain"-programs compete in a virtual world.
The first path is being attempted by
Ben Goertzel,
the second is championed by
Ray Kurzweil,
and the third is the path chosen by alifegames.com.
We figure that our project will have to run for a very long time before it
comes even close to answering the big questions. In the first phase, it
will be nothing but a testbed for the development of alife-based AI for use
in computer games, first for role playing games, then, as we make the
environment more complex, for the next generation of real-time-strategy games.
Our project is based on two paradigms:
- Situated Action: We let our brains evolve in an environment they
directly interact with - we make no attempt to model their environment and
let them interact with the model;
- Society of Mind: We allow brains that consist of numerous interacting
sub-modules, which can be of quite distinct types - for every task we
try to find the simplest architecture that can solve it.
The environment our evolving brains interact with is initially simple,
but will be made more complex step by step until eventually
it will become a good model of a 3D-virtual world. If the brains we evolve
are downloaded into suitable robot bodies, they should then be able to interact with
our natural world.
Our first version of a virtual world will consists of
DungeonMaker-generated dungeons and
labyrinths, which are inhabited by MOBs (mobile objects) which do their own
thing, but will cooperate to hunt players who kill their brethren and
steal their treasure. The game will feature pluggable brains written in a
scripting language, so that users of the program can easily try out
their own MOB-AI. If you have thoughts
that might help this project along, please post them on
the list.
Once the first version of the program is up, we will also look for coders and artists.
If you might be interested, you should monitor one of our mailing lists and step forward
when a task comes up that you are interested in.