Commercial interruption: the best thing about ALife is Santa Fe - artificial life, Santa Fe Institute

RELease 1.0, April 28, 1992 by Esther Dyson

The Santa Fe Institute was founded in 1984 in a sort of spontaneous generation that involved George Cowan, its first president, and six of his colleagues from the Los Alamos National Laboratory. Other early figures included man-about-physics Murray Gell-Mann. They sat around and talked about the idea, then held a workshop to test the level of interest in an interdisciplinary institute devoted to the study of complex, adaptive systems. The idea survived, and the group has attracted funds to live year to year (close to $3 million annually) but no permanent endowment.

With a science board of 50 and a trustees' board of 40 (including Esther Dyson), the Institute by default has become the premier organization devoted to the field. It is looking for a permanent chairman; the acting chairman is Jim Pelkey, a venture capitalist and consultant (and former ceo of Sorcim) who moved to Santa Fe two years ago. The Institute has a resident staff of about 25 including 10 scientists, and a visiting faculty ranging from 15 to 40 scientists from month to month. Much of its work is carried out through workshops, symposiums and other meetings, where scholars meet others from different fields to share ideas. Its major fields of study are economics (which somehow has the easier time getting funded) and a number of harder-to-categorize projects such as those described here. One way or another, all depend on computers.

SFI's major purpose is to keep on experimenting. SFI meetings are fun: In most scientific discussions, when you ask questions, the answer is,

"We don't know because ... there's some serious impediment to finding out. In our own business of software, it's usually, No one's done it yet, and it will take a few years to happen." Or, They're working on it, but it's late." At the Institute, the answer is usually, Well, let's try that."

A list of some people we ran into on our last visit gives a flavor of the place: Lev Zhivotovsky, a Moscow geneticist; Carl Djerassi, founder of Syntex; Bob Maxfield, co-founder of Rolm Corporation; Chris Langton, prime mover of artificial life from Los Alamos; Melanie Mitchell, resident director of the Institute's adaptive computation program, who has worked with John Holland and Doug Hofstadter at the University of Michigan on a variety of AI projects; Danny Hillis, founder of Thinking Machines Corporation and designer of its Connection Machines; Doug Carlston, founder of Broderbund; Marc Feldman, a Stanford population geneticist; Tom Ray and John Holland, described in this issue; Will Wright, creator of SimCity and other games; Stuart Kauffman (cousin of Objective Software's Rich Melmon), MacArthur Fellow and theoretical biologist; David Liddle, now setting up Interval Research; and Doyne Farmer, founder of Prediction Company, which will use adaptive computation to direct trading activities in financial markets.

We were there for the Institute's recent founding workshop on adaptive computation. The purpose was both to share research, and for the Institute to survey current work in order to shape its adaptive computation program for the next few years.

NATURAL HISTORY OF ARTIFICIAL LIFE

Artificial life gained visibility with the first conference on artificial life in September of 1987, sponsored by the Santa Fe Institute and Los Alamos National Lab's Center for Nonlinear Studies. (Call it Los Alamos Nonlinear Labs?) That conference attracted about 150 people; the second one, in 1990, attracted 350; and the next one, this June 15 to 19, will be limited to 500. Also in June several books on ALife will appear, including one by Hackers" author Steve Levy. In short, the field is about to be very fashionable. Although its successes so far are mostly demonstrations, experiments and projects, the field has caught the attention of (and funding from) the likes of IBM, Apple, DEC, Sun - and Citibank, financial house O'Connor & Associates and others.

The sociology of the field is fascinating in itself: Nobel-winners, biologists, sociologists, physicists, economists and psychologists, many of them Nobel-winners, come into close contact with each other as well as with social butterflies, science-fiction writers and science salon-keepers. Indeed, the field gets much of its vitality from this cross-pollination (to use a "life" metaphor loosely). Moreover, many leading complex-systems people aren't quite respectable in their own fields: They're considered strange, weird, not-mainstream and worse by their colleagues. At the Institute, by contrast, they get respect, but there's the opposite problem: The physicists don't always understand biology as well as they think they do, and vice versa. One of the Institute's major contributions is exposing the different disciplines to each other.

Hillis on holiday

Another early contribution to the field was Danny Hillis's work on generating sort algorithms (for 16 elements) by evolution, which he somehow fit into his role as founder and chief scientist of Thinking Machines Corporation. (See the PC Forum proceedings for 1990.) Using a genetic algorithm and a Connection Machine, he started with a set of compares as raw material. The program created new sequences to be tested against random sequences of 16 numbers, and the best scores won each round. The better-than-average scorers were then split and recombined at arbitrary points, and the population improved gradually. The resulting program was quite compact, involving a sequence of compares, but it did the job in 65 steps, almost matching the world record of 60. This is a classic example of optimization. It took a few hours on a Connection Machine.