Maximize
Bookmark

VX Heaven

Library Collection Sources Engines Constructors Simulators Utilities Links Forum

Interview with Tom Ray

SPTH
Ready Rangers Liberation Front [7]
June 2006

[Back to index] [Comments]

About Tom Ray

Dr. Thomas S. Ray is a professor of Zoology at the University of Oklahoma, with an adjunct appointment as Professor of Computer Science. He received his Masters and Doctorate in Biology from Harvard University, specializing in plant ecology.

Tom Ray is a well-known scientist in the field of artificial life and evolution. In the 1990s he has developed "Tierra" and "Virtual Life". Beside of that, he has also been deeply involved in rain forest conservation in Costa Rica.

His homepage: http://life.ou.edu/

About Tierra

Tierra is a computer simulation developed in the early 1990s in which computer programs compete for central processor unit (CPU) time and access to main memory.

The computer programs in Tierra are evolvable and can mutate, self-replicate and recombine. Tierra is a frequently cited example of an artificial life model; in the metaphor of the Tierra, the evolvable computer programs can be considered as digital organisms which compete for energy (CPU time) and resources (main memory).

(http://en.wikipedia.org/wiki/Tierra_%28computer_simulation%29 and at Tom Ray's homepage)

The Interview

Why have you started to do researching in such an unusual topic as "artificial life"?

I feel that I answered this question in my essay:

Ray, T. S. 1993. How I created life in a virtual universe.

http://life.ou.edu/pubs/nathist/

After you have introduced [email protected], which has terrifically increased the size and speed of Tierra, have there been any significant successes (social behaviour, increase of complexity, ...)?

Although I have released the source code for the network version of Tierra, I have never fostered a [email protected] project. The reason is that I am not convinced that running a really large network would produce the desired result, a virtual Cambrian explosion.

In the real world there are hundreds or thousands of chemical reactions, which are equally to computer functions in Tierra. But the amount of functions in Tierra is limited in comparison to Operation Systems or BIOS functions. Would be a more complex system more suitable for evolution?

I don't think your analogy is appropriate:

	chemical reactions = functions in Tierra

I think I might equate chemical reactions the actions of algorithms, or code fragments, of which there are a limitless number.

Soon in Tierra parasites have been formed, later cells with immunity, which has been a highly interesting development. What has been the most impressive development in Tierra in your opinion?

I think the ecological/evolutionary dynamic was the most impressive result in Tierra.

Sexuality is a very successful way of recombining the genomes of creatures in biology. Is this system also successful in artificial evolution?

Yes. I don't use it much in Tierra, but it is used a lot in genetic algorithms and genetic programming.

In some of your publications you mentioned the phrase "Cambrian explosion" in computers. In the history of life this happened more than 1 billion years after the development of first forms of life. In addition, earth had more resources than networks like Tierra - even more than the whole worldwide network. Is there still a reason for believing that this will happen in computers much faster?

The Cambrian explosion happened when organisms were prepared to explore the diversity of the multi-cellular developmental process. It took about three billion years to reach that point. However, in an artificial system, we can attempt to engineer the conditions from which that exploration is possible.

What has prepared biological organisms for that important step of evolution?

Some say that it was necessary for the concentration of oxygen in the atmosphere to reach a level high enough to support large multicellular organisms. But I also think that it had to await the evolution of a flexible genetic system for directing the developmental process, such that mutations in the developmental controls could produce changes in body plan.

How do you want to engineer this condition in an artificial system?

The network Tierra experiment was not about scale, it was about engineering a system poised on the threshold of a digital Cambrian. This is discussed in papers such as:

Ray, T. S. 1998. Selecting Naturally for Differentiation: preliminary evolutionary results. Complexity, 3(5): 25-33. John Wiley & Sons, Inc.

Ray, T. S. and Joseph Hart. 1998 Evolution of Differentiated Multi-threaded Digital Organisms. In: Artificial Life VI proceedings, C. Adami, R. K. Belew, H. Kitano, and C. E. Taylor [eds.], 295-304. The MIT Press, Cambridge.

The multicellular starting organisms in network Tierra were differentiated at the most primitive level: two cell types. The object of the experiment was to observe an increase in the number of cell types.

You have compared today's software with biological cells, and wrote that even our most complex software (such as operation systems including multi-threading) are just Protozoa (single-celled organism). What could be the "Cambrian explosion" in computers and what do you expect from it?

I discuss this in a paper that I think you have read:

http://life.ou.edu/pubs/reserves/node1.html

I point out that we probably can not imagine what such digital organisms might be like.

Are there conditions, which influence the speed of evolution in a positive ways - in biological and artificial systems?

In biological systems, adaptive radiations occur when species invade ecological voids (such as the colonization of a newly formed archipelago, e.g. Galapagos) or when they evolve new morphological novelties (such as multicellularity, e.g. Cambrian explosion; or jointed appendages, e.g. diversification of insects). Artificial systems unfortunately are orders of magnitude simpler that biological systems that exhibit adaptive radiations. Many studies in genetic algorithms or evolutionary computation compare rates of evolution under different conditions. But this is all trivial compared to what happens in biological systems.

The whole theory of evolution relies on randomness. New let's change the view: Is there also a way of mutation, where artificial creatures try to improve their own code by calculate a best-case-scene, and what is your opinion about that?

This kind of reverse engineering is not practical for a variety of reasons. Organisms don't have any means of forseeing what kinds of changes are needed for adaptation. Also, if it were known what kind of adaptation is needed, it is generally not possible to know what kind of mutations would generate such an adaptation. We don't have a general understanding of the mapping from genotype to phenotype.

In 1986 Fred Cohen wrote in "Computer Viruses - Theory and Experiments" about the possibility of accidental development of artificial life due to software bugs. Could this thought come true in our modern world, which is "dominated" by computers?

David Ackley gave a fascinating lecture about "real" AL several years ago, in which he claimed that all the critters living on the net (viruses, worms, etc) are in fact artificial life. This is one view, but see my answer to the next question.

Several sources count computer viruses as first artificial life. What is your opinion about that statement and what do you think about harmless computer viruses, which have just been created for the purpose of existing?

This question forces us to take a position on defining life. Computer viruses, worms, etc exhibit self-replication, an important quality of life. However, I consider evolution to be both the defining property and the creative process of life. Evolution is the missing element in the kind of "real" AL that Ackley talks about.

What do you consider as "computer virus"?

I think this is adequately answered here:

http://en.wikipedia.org/wiki/Computer_virus

How would you describe a program with self-replication- and evolution-capability? Computer virus or artificial life?

I think if it both self-replicates and evolves, we can consider it to be artificial life.

Beside of self-replication and evolution, what else do you consider as conditions of life?

I think self-replication and evolution are what matters, but many people will produce a long list of characters, including such things as metabolism and repair.

Metabolism is biochemical process mainly for getting energy and for anabolism - both is not needed for an artificial organism. What else could "artificial metabolism" be?

We can only make analogies between organic and digital life. These analogies may or may not be meaningful. A digital analog of metabolism might just be the processing of instructions and manipulation of data.

In Tierra, phenotype and genotype of the creatures are the same. What could be the phenotype of an artificial creature, if we do not count simulated "earth-like" systems as Karl Sims' "Virtual Life"?

I think it is a mistake to see the phenotype and genotype as the same in Tierra. The genotype is the bit sequence or instruction sequence of the genome. The phenotype is what happens when those instructions are executed.

In John Conway's "Game of Life" 2-dimensional "pixel-creatures" behave as they would life. What do you think about this mathematical simulation - do you consider its creatures as alive?

I enjoyed playing with it as a kid, when it was published in Scientific American. Today I don't find it very compelling. The entities don't have a genetic representations, thus there is really no heredity, so there can not be a basis for evolution, and there is no evolution in the biological sense. I don't consider them to be alive.

In "The Universe in a Nutshell" Stephen Hawking explains that (biological) life is just possible at universe with more then 2 dimensions. What do you think about that idea, and is there something equal to dimensions in the computer?

This may be true in the context of the material universe. However, I don't consider it to be true in the digital universe, where dimensionality has quite different qualities. I discussed dimensionality here:

http://life.ou.edu/pubs/zen/node18.html

By the way, I urge you to read the zen paper:

http://life.ou.edu/pubs/zen/

It may be my best paper.

Are you interested in any other artificial life or evolution projects?

I like Karl Sims evolved virtual creatures very much, and I created a version of it that runs on the PC: http://life.ou.edu/VirtualLife/

What do you expect about artificial life and about Tierra in future?

Open ended evolution, and large evolutionary increases in complexity remain challenges in AL. I don't know what the prospects are of overcoming these challenges.

What are your current researches, and what are your future plans about that topic?

I am currently working in the area of psychopharmacology, trying to understand the chemical architecture of the mind.

[Back to index] [Comments]
By accessing, viewing, downloading or otherwise using this content you agree to be bound by the Terms of Use! vxheaven.org aka vx.netlux.org
deenesitfrplruua