What is Artificial Life?
Virtual Pet Research
Joseph Bates is the Principal Research Computer Scientist at Carnegie
Mellon University (CMU). Professor Bates studies artificial intelligence,
interactive art and interactive entertainment. At CMU he leads the Oz project,
an interdisciplinary effort to create believable interactive characters
and emotionally powerful interactive stories.
Fujitsu Laboratories and the Oz group have collaborated on shared research interests since 1990. Some results of this research appear in Teo.
Polson Enterprises Comment
Project OZ at Carnegie Mellon Univ.
Q: Define "Artificial Life" ("A-Life") and describe what makes Fujitsu's Teo technology Artificial Life.
A: Artificial Life is the attempt to build systems that exhibit some of the key properties of life, such as adaptability and robustness in complex environments. It is different from "Artificial Intelligence" (AI), which has traditionally focused on cognitive and intellectual abilities of human beings.
There are many aspects of Artificial Life and many approaches to achieving it. The technology of the Teo Project allows artists and engineers to create creatures that rely heavily on perception, instead of thought, and thus are very responsive to their environment (which includes human beings in front of the monitor). Teo can probably be best thought of as a combination of A-Life and AI technologies, carefully chosen and combined by artists, to produce believable autonomous creatures.
Q: How does Fujitsu's A-life technology constitute a breakthrough?
A: I think the way to say this is that technology from the most advanced AI research centers around the world is making its first appearance in the Teo product line. These are technologies that are usually seen only in the research labs, but Teo is making them available to consumers.
Q: The Teo Project is presented as the first consumer product incorporating A-Life. Is this accurate from your vantage point?
A: I think Maxis has marketed products that might be called A-Life, such as El-Fish. But I think Teo has a more sophisticated model of mind inside each creature, which came from Teo's and Oz's [a Carnegie Mellon University project] AI history. This richer model allows Teo creatures to seem more like mammals, instead of fish.
Q: How accurate is the presentation of the Teo Project as an upwardly scaleable technology with increases in interactivity and functionality?
A: I believe this is accurate. We are constantly working to improve the technology and art of the mind and body. We should see continuous improvement over many years (or decades). The underlying technology may change occasionally, but the user will see ever improving personalities in the creatures, more interactivity, and so on.
Q: What new A-Life functionalities will be forthcoming from the Teo Project group? When can we expect them?
A: Well, I wouldn't call the improvements purely A-Life. For example, a creature that can use language (for instance, generating and somewhat understanding text) will be a significant improvement. This is usually considered beyond the scope of A-Life -- it is closer to AI. So I think we will see a variety of improvements in emotion, language use, social abilities, perception, body motion, intelligence, and so on.
I'm not sure when the Teo group will release these things -- a mixture of AI, A-Life, and art. But generally speaking many of these improvements could happen gradually over the next five years.
Q: How does Fujitsu's A-Life technology differ from or relate to other hard, wet and soft AI, as well as other A-Life efforts such as the work emerging from the Santa Fe Institute and others?
A: Again, Santa Fe is doing very pure A-Life. The result is creatures that are very adaptive and robust, but of very limited capability. Teo, Fujitsu and Oz are trying to build entire creatures that are interesting to humans. They must be emotional, social, and what the artists call believable. These artistic qualities may not be important in the AI and A-Life labs, but they are crucial to success in the consumer market.
Our research and development efforts use any available technology (hard, soft, AI, A-Life) from the fundamental research centers. The key to our work is letting artists guide the use (and even creation) of this technology, and then bringing it to consumers rapidly.
Q: How does A-Life and particularly Fujitsu's A-Life technology fit in with the emerging intelligent agent technology research being conducted at Massachusetts Institute of Technology and other institutions?
A: Most intelligent agent researchers are striving for various forms of intelligence. We believe people care much more about emotion and believability. So we use and develop similar technologies, but we aim for artistic goals, instead of traditional scientific goals.
In the future, a creature that combines intelligence with believability could be very useful. Many researchers are developing the intelligence part. We will be one of the few who know how to achieve believability.
Q: How does Fujitsu's A-Life technology relate historically to the trajectory of AI development from expert systems, fuzzy logic, neural net, etc. to where Fujitsu is today? What is the historical linkage, legacy and breakthrough here?
A: Good question. Some of the original goals of AI were expressed very clearly by a past president of the American Association for Artificial Intelligence. He said that he began as an AI researcher because he had "a dream to build a computer friend - one that could understand, act autonomously, think, learn, enjoy, hate - which liked to walk and play ping-pong, especially with me."
This old dream of AI really is a blend of desires for intelligent creatures and for believable creatures. But for decades most AI researchers have studied only intelligence. So Oz and Teo have returned to study the original goals that drove the founders of the field of AI. We have concluded the artists understand best what are the key properties of such creatures, and we must listen to them as we pursue AI's original dream.
This is what Oz has been doing - trying to build believable autonomous creatures, not intelligent autonomous creatures. And now Teo has brought shared Fujitsu/Oz initial research results to the consumer.
It is not so much a breakthrough of technology but a breakthrough in rethinking the old goals and old dream. Taking the arts very seriously as a contributor to AI is a breakthrough in research methodology.
Q: Is Fin Fin the first true artificial life form?
A: It is not possible to define life. If you want to be scientifically accurate, you cannot call Fin Fin the first true artificial life form, because we do not know how to define life. For many decades there have been things that people could choose to call A-Life. You might call Fin Fin the first commercially available believable autonomous character. You can certainly refer to certain films, books or amusement parks as places where believable characters exist, but they are not autonomous or interactive. Fin Fin is the first artistically created believable character that also thinks for itself.
Return to Virtual Pet Home Page