EDGE 20

EDGE 20 — June 23, 1997

DIGERATI

"INTENTIONAL PROGRAMMING"
A Talk with Charles Simonyi ("The WYSIWYG")

The "first law" of intentional programming says: For every abstraction one should be able to define an equal and opposite "concretion". So repeated abstraction or parameterization need no longer create "Turing tarpits" where everything eventually grinds to a halt due to the overhead introduced by the layers. In IP, the enzymes associated by the abstractions can optimize out the overhead, based on the enzymes' domain specific knowledge. The overhead associated with abstraction has always been the bane of the very-high-level languages in the past.

THE REALITY CLUB

Alun Anderson, John Maddox, Lee Smolin

Arnold Trehub & Steven Quartz on "Organs of Computation"

(11,002 words)

John Brockman, Editor and Publisher | Kip Parent, Webmaster

DIGERATI

INTENTIONAL PROGRAMMING
A Talk with Charles Simonyi ("The WYSIWYG")

During the 1970s at Xerox PARC, Charles Simonyi led a team of programmers in the development of Bravo, the first WYSIWYG (what you see is what you get) word-processing editor. Bravo was a fundamental departure from the way information was previously displayed and organized and it was part of PARC's contribution that changed the face of computing and ultimately led to personal computing.

Simonyi, born in Budapest, Hungary, holds a bachelor of science degree in engineering mathematics from the University of California at Berkeley and a doctorate in computer science from Stanford University. He worked for the Xerox Palo Alto Research Center from 1972-80 and joined Microsoft in 1981 to start the development of microcomputer application programs. He hired and managed teams who developed Microsoft Multiplan, Word, Excel, and other applications. In 1991, he moved to Microsoft Research where he has been focusing on Intentional Programming. He is generally thought of as one of the most talented programmers at Microsoft.

Dr. Simonyi, whose long career has made him independently wealthy, has endowed the two chairs: the Charles Simonyi Professorship For The Understanding Of Science at Oxford University which is held by the evolutionary biologist Richard Dawkins; and the Charles Simonyi Professorship in Theoretical Physics at the Institute for Advanced Study.

John Markoff, writing in The New York Times (12 Nov 1990), relates the following anecdote: "He enjoys taking visitors to the machine shop in the basement of his new home, complete with lathe and drill press. 'In Hungary,' he said, 'they told us that the workers would never own the
means of production.'"-

JB

CHARLES SIMONYI is Chief Architect, Microsoft Corporation.

INTENTIONAL PROGRAMMING
A Talk with Charles Simonyi ("The WYSIWYG")

JB: What's new, Charles?

SIMONYI: I have been working on what we call "intentional programming." It's very exciting. It has to do with professional programming, so it's kind of hard to get into the details. It also relates to the work of evolutionary biologist Richard Dawkins in a fairly direct way.

We are trying to create an ecology of abstractions. Abstraction is really the most powerful tool that we have in thinking about problems. An abstraction is a single thing, yet if it is a good one, it can have many applications, even an infinite number of them. So an abstraction may be looked at from one side as a compression of many instances into one generality or from the other side as a special purpose power tool that yields the solution for many problems. If one could attach a dollar sign to this power, the economies would be amazing: rivaling that of chips or application software itself.

Programming languages are really just vehicles to supply abstractions to programmers. People think of programming languages as being good or bad for a given purpose, but they are really criticizing the abstractions that a language embodies. The progress in programming languages has been incredibly slow because new programming languages are difficult to create and even more difficult to get adopted. When you have a new programming language, the users have to rewrite their legacy code and change their skills to accommodate the language. So, basically, new programming languages can come about only when there is an independent revolution that justifies the waste of the legacy, such as Unix which gave rise to C, or the Web which gave rise to Java. Yet it's not the languages that are of value, but only the abstractions that the languages carry.

It's very much like Dawkins' idea that it's the genes, not the individuals, that are important in evolution. And, in fact, what's being reproduced are the genes, not individuals. Otherwise, how would we have worker bees and so on. We are doing the same thing; it's abstractions that matter, not languages. It's just that we don't think of abstractions without languages, because languages used to be the only carriers for abstractions. But if you could create an ecology in which an abstraction could survive independent of everything else, then you would see a much more rapid evolution for abstractions, and you would witness the evolution of much more capable abstractions.

To enable the ecology, all you have to do is make the abstractions completely self-describing, so that an abstraction will carry all of its description, both of how it looks and of what it does. It's called intentional programming because the abstractions really represent the programmers' original computational intent. And that's what the important invariant is, everything else of how something looks or how something is implemented, these are things that should evolve and should be improved so they can change. What you want to maintain invariantly is the computational intent as separated from implementation detail.

JB: It sounds biological in nature.

SIMONYI: Yes, we are using a lot of biological metaphors. We call our transformations, for example, enzymes. It's just that biology, and all the sciences of complexity, are making big forward strides, and it's just a matter of using as many of the metaphors as one can.

JB: But it is still a programming language, isn't it?

SIMONYI: Absolutely not. Intentional Programming relates to a Programming Language as a powerset relates to a set. It is strictly greater, there cannot be any isomorphism between the two. IP programs are encoded in a tree-like data structure where each node also has a graph like pointer to the definition of the intention the node is an instance of. Every node can have arbitrary nodes underneath it, that is nodes can be parameterized arbitrarily. The semantics of intentions are described by the tree transformations which convert the instances into primitive intentions from which native or interpreted code can be generated by standard means. The looks of intentions are also defined by arbitrary computation which serves no purpose other than to ease interaction with the programmer. So names and looks — what used to be called "syntax" — will have no effect on the computation and may be changed arbitrarily as programming styles and programmers' needs evolve. Knuth's dream of "literate programming" will become practical and I expect a great deal of visual richness to also emerge in the programs.

JB: Isn't this just another meteor that wipes out legacy to make room for evolution?

SIMONYI: Luckily that is not the case. Intentions can be defined for all features of all legacy languages, so legacy code can be imported into IP without loss of information or functionality. Once in IP, the process of incremental, continuous improvement can begin and the lifetime of the legacy code will be limited only by its usefulness, not by the means used to encode it.

JB: Do you foresee structural changes in the industry as a result of this?

SIMONYI: It will be very exciting. The personal computer industry has enabled evolution in the platforms. Out of the Cambrian Explosion of the early eighties there emerged a few dominant architectures, the Windows family being the most popular of them. There is incredible variety in terms of peripherals, applications, networking, form factors, performances all the result of evolution. I foresee a similar progression in the realm of abstractions. Once everybody with a $5K machine and good programming skills is empowered to create and publish abstractions for which any one of the tens of millions of programmers will be potential customers, there will be a tremendous explosion of creativity. Many of the early new abstractions will be addressing the same easily accessible niches, such as collections and maps of course, so a shakeout will be inevitable. Then the creative energies will be channeled to the myriad domains: software sharing, user interfaces, graphics, accounting, animal husbandry, whatever. Each of these areas will benefit from domain-specific abstractions and optimizations which in turn will improve the quantity and quality of application software in those domains. There will be more shareable software artifacts, thanks to IP's ability to parameterize any abstraction further with any kinds of parameters.

The "first law" of intentional programming says: For every abstraction one should be able to define an equal and opposite "concretion". So repeated abstraction or parameterization need no longer create "Turing tarpits" where everything eventually grinds to a halt due to the overhead introduced by the layers. In IP, the enzymes associated by the abstractions can optimize out the overhead, based on the enzymes' domain specific knowledge. The overhead associated with abstraction has always been the bane of the very-high-level languages in the past.

There will be markets for flexible artifacts, abstractions in all domains, and different implementations for those abstractions. These in turn will improve tremendously the quality and availability of application software.

Once one looks at abstraction as a commodity, the standard rules of business can be applied. For example, one should be able to invest in the manufacturing process in order to make the product more competitive. Or in the software arena one should be able to elaborate the definition of an abstraction in order make its application more effective: simpler, faster, more flexible. Conventional programming languages completely ignored this elementary rule: declarations and references were all designed with the same high-falutin principles in mind: orthogonality, cleanliness, what have you. It's as if we used the same standards, materials, and budgets for the factory and the retail store, or for the machine tools and for the product. Nobody in business would behave in such an idiotic fashion. In IP one can associate an arbitrary computation with any definition, so the opportunities for making investments in definitions are limitless.

JB: I've been hearing your name since the mid-seventies. It seems like you've been a player in almost every epic of personal computing.

SIMONYI: I've been incredibly lucky, in a strange way. In the U.S., computers that operated with vacuum tubes were obsolete in the late 50's, whereas in Hungary, where I grew up, they were in use. It was a time-warp. Also, I started working with computers at a young age. When the personal computer revolution came about much later, the people in the U.S. that had worked with tube computers were long retired, if not dead, while I was really in the prime of my career. So starting very young and starting in a time-warp gave me this double benefit that I think very few people had. It was very unusual, at least in Hungary, to start that young, but if you look at it today, you know that computer programming is not difficult. Or rather, the kind of computer programming that was done in the 60's is really child's play. It's just that at that time that secret was well hidden, so people were very worried about letting me close to very expensive machines. But I had a certain pull through my dad, who was a professor of electrical engineering, and I made myself useful by working for free, which was also a kind of an unknown notion, but I had this intuitive idea of looking at it as an investment.

This was in Hungary, in 1965, when I was 16 years old. I learned a lot then. In a period of three years I traversed three generations of computers: the first generation in Hungary, then a year and a half in Denmark on a very typical second generation machine in Copenhagen. Then I proceeded to Berkeley where I wound up in a computer center on a CDC 6400, which was a fantastic third generation machine.

JB: How did you get out of Hungary?

SIMONYI: I went to a lot of subterfuge to get out, to be sure. I got out legally; but it was illegal not to return. The way I got out was that I finished high school one year before it was expected, which was a unique feat at the time, in terms of actually getting permission and go through with it. People there were living in a very fearful and lock-step way, and just to do something unusual was a big deal. So when I got out of high school I was 17, underage, so the military couldn't touch me. I also secured an invitation to work at Regnecentralen in Copenhagen, where one of the first Algol compilers was developed. At that time university people had deferments, so the military were in a quandary. If I were to go to university in Hungary, then I would have been completely out of their reach. Whereas if I had spent the year by going to this Danish institute, for example, then they could catch me on the rebound. So they took the lesser of two evils, and they let me out.

JB: Did your father leave with you?

SIMONYI: No, I left alone. He had many political problems and later suffered because of my defection, but we had already taken that into consideration. It worked out for the best in the end, and he would have been very unhappy if I had been on his side having the same problems as he did. He didn't want to leave his country for many reasons, and I think he was egging me on to leave. I mean, now I can say freely that he was encouraging me to get out.

JB: What happened at Berkeley?

SIMONYI: I got there when I was 18, and I was kind of a starving student. Basically, I had a lot of problems with the immigration people, because nobody had been shooting at me at the Hungarian border. I was just a normal student, except a student whose passport seemed to expire every minute. Though I had plenty of offers to be a programmer, they were pretty strict about taking up employment, which I thought was very strange in the land of the free. Also, you couldn't get scholarships as a foreign student, so I was pretty much living without visible means of support.

I worked for the Computer Center first and met Butler Lampson and did some jobs for him. He and some other professors started a company called Berkeley Computer Corporation, and they invited me to work for them, and that's when I first received a stock option. It wasn't worth anything in the end, but it's a funny story I haven't told before.

Sometimes I was an outstanding student and sometimes I was a terrible student, depending on if I had money or if I had to work or whatever. Also, I had no incentive to get good grades; I just wanted to get an education. I was completely on my own; I paid for it myself; I viewed myself as the customer, and a grade was just some stupid rule that the university had. So I optimized my grades just so they won't throw me out. Anyway, the Dean talked to me and said, well, Mr. Simonyi, you were doing so well and are now doing so poorly; what's the reason? Can we help you? You can share anything with us, tell us what it is. Is it drugs, is it grass, acid, or mescaline? I smiled at him and said, I think it's a stock option. He said, well in that case we can't help you.

Berkeley Computer was really an offshoot of Project Genie, which was funded by ARPA, and Bob Taylor was doing the funding. When Berkeley Computer went bankrupt, the core people were hired by Bob Taylor who was working for Xerox by then. This is how I got into Xerox PARC. I still didn't have my Bachelors degree. With all this skimming the bottom of the expulsion curve, it took me five years to get the degree.

At PARC I had a number of different projects. Then the Alto came into being — the first personal computer — and we had this fantastic capability that was so evident. The most interesting thing: when you see a capability that kind of blows you away, and you know that this is going to be the biggest thing, but then some people don't see it. So it's not like Alto was the only project at PARC; it was just one of a number of similar projects that was fighting for resources. A resource is more than just dollars, it's all forms of attention, attention of the key people and so on.

One day I saw some pieces of paper on Butler's desk, and I asked him what it was, and he said, Oh, it's a sketch for a text editor, we need that for the Alto, and I said, well, can I take a look at them? He said yes, there's nobody working on it. So I took it and decided to make it happen, because it looked very sweet.

We had to create, again, a subterfuge to make it happen. I had to do some experiments on programmer productivity, for my Ph.D. thesis. The first experiment was called Alpha, the second experiment was Bravo. That's how the first WYSIWYG editor was called Bravo, and it was funded, in a way, as an experiment for part of my thesis.

My thesis was not about WYSIWYG. The thesis had some philosophical parts and some measurement parts. The measurement parts are pretty useless, the philosophical part was quite good, and it served us well later in the early days of Microsoft. It had to do with organizing teams, looking at projects, naming conventions and evolving techniques.

Meanwhile, of course, WYSIWYG was born. Once the Bravo editor and the other component of the "office of the future" were operational, it created a fair amount of attention, and a lot of VIPs came to look at what PARC was coming up with. The name WYSIWYG came about during a visit from Citibank representatives. We had a demo showing how we could display a memo with nice fonts, and specifically the Xerox logo in its specific Xerox font, on screen, and then send it through the Ethernet and print it out on the laser printer. So we printed what we had created on the screen onto transparent slide stock. Part of the demo was to push the button to print and then we held the printed version up, in front of the screen, so you could see through the transparent stock that the two were identical. Actually they weren't exactly identical, but they were close enough. It was pretty impressive.

One of these visitors said, "I see, what you see is what you get." Which was of course, you must remember, the Flip Wilson tag-line from Laugh-In, which was a big TV hit at the time. I think he was doing a female impersonation. What you see is what you get. Well, that's the first time I heard it used around the system, which was the first incorporation of that idea, so somehow the term WYSIWYG must have spread from that event.

JB: How did developing the WYSIWYG word processor lead you to Microsoft. Or, rather, was Microsoft then in existence?

SIMONYI: Microsoft might well have been founded on the very day we gave that demo to Citibank in 1975.

At that time we already had a Mac, with a bigger screen than Mac, with a mouse and so on. The Alto was a very, very serious machine. It cost fifty thousand bucks; the printer cost two hundred thousand bucks, in 1975 dollars. Gosh, I remember thinking that maybe one day the drugstore at the corner might have one of those machines, and then it might be shared by the whole block, or a whole area in a city. Now I have several of them at home.

Microsoft began at that time by doing Microsoft Basic. I started to hear about microcomputers in E78, E79, and it sounded like a kids toy. I recall that Larry Tesler at PARC had a Commodore 64 in his office, and we sometimes went there to smile at it. I certainly never took it seriously.

Eventually I started to become deeply unhappy because Xerox seemed to be treating these ideas in an incompetent fashion. My fear was that I would be missing out because I was allied with Xerox, and that the world would be missing out because they were not going to get what was possible. It wasn't just the Xerox marketing or management organization, but also the technical organizations, that share a lot of the blame if it should be called blame. Perhaps we should just think of it as evolution in action.

The failure of Xerox saved me from a fate worse than death, which would have been not sharing in the success. If Xerox had been successful, I would have gotten a thousand dollar bonus, and that would have been it. And I would have felt a little bit dumb.

But I didn't see the future until I saw Visicalc running on an Apple II. That was a capability that we didn't have. I thought Xerox suffered from a disease we call "biggerism," which is the bigger-the-better type of engineering mentality. And it always escalates and compounds, and it results in very complicated and very expensive machines, which is very, very risky, because it's very difficult to change or to tune to the market, or even discover what the market wants.

I saw this nimble machine, the Apple II, providing both some functionality which we at Xerox did not possess, and also having an incredible biological advantage

JB: Then what?

SIMONYI: I met Bill Gates, and I clicked with him right away, very quickly in a very intense way. He was still very, very young, in his early 20's. This was in November of 1980. But the scope of his vision was extraordinary for the time, including his ideas about the graphical user interface. We had a discussion, and I came back a couple of weeks later with a summary of the discussion. Bill saw Microsoft becoming the leading producer of microcomputer software, worldwide. We wanted a worldwide, international market, and to be a leading producer with a full offering of operating systems, applications, languages, and consumer products.

It was easy for me to believe in the importance of applications and graphical user interface because of my experience at Xerox. It was amazing to see it coming from Bill with an equal intensity, when he hadn't seen all that stuff, certainly not with the same intimacy as I had. Furthermore, I realized that he actually had the wherewithal to deliver it. It was interesting to look at a company like Xerox, with a hundred thousand people and billions of dollars, and realize that the success of your project depends on having the right two people that you want to hire, who may not fit into the corporate structure. And then you realize that this single guy can hire anybody he wants to! Bill just said, hire two people, or hire five people. What do you need? Do you need rooms? Do you need chairs? Yeah! We can do that. Computers? Yes. You need this, you need that. Sure. We were talking about only a few hundred thousand dollars which could make a difference, we weren't talking about a billion.

Bill did spend a lot of money on one thing a Xerox Star. We got one of these biggered, enormous, expensive machines, but it had the germ of the right idea in it. And we just wanted everybody in the organization to get used to the desktop and to the mouse and to pointing and to what's possible. And if it's not perfect, that's fine. We didn't want to use operationally; we used it for education of the people.

I described myself in an interview as the messenger RNA of the Parc virus. I never thought the journalist would use it, because at the time nobody was talking about viruses, about DNA, let alone RNA, let alone messenger RNA, let alone getting the metaphor. But it was used.

It was the biggest thing in my life, certainly, joining Microsoft and getting involved in the tremendous energy of those years. Probably one of the most important things that we did was the hiring. That's one of the enabling factors of growth, and I think we did a super job in hiring. Many those people are still with us, and many of them are in very high positions in the company. But, more than for any of our competitors, they formed a very responsive, very efficient programming organization.

That was key, because we did have problems. In applications, we had to be able to do spreadsheets, we had to do word processing, we had to do the databases. It was a no-brainer to know that. We did a fairly good job in spreadsheets. We were competing very effectively against Visicalc using a strategy that is very much like Java today; it was a platform independent, interpretive, bytecoded system, that enabled us at one point to run on 100 different platforms, so we could fund the international organization and get acquainted with localization problems, and all those things. Actually, Multiplan, our spreadsheet, remained very popular in Europe, for much longer than in the States. What happened in the States was that Lotus 1-2-3, wiped us out. So that was kind of difficult, but it was our fault. We were betting on the wrong horse — the mixed market with a variety of systems, instead of the right horse, which happened to be also ourselves, namely MS-DOS.

JB: The software war du jour.

SIMONYI: Out of this debacle came Excel later on. I think that competition is very important. It obviously creates again, comes from biology much better results. If you look at evolution, much of evolution is not in response to the environment, it's response to other flora and fauna in the environment. So we need that, and it's sad when the opponent doesn't put up a good fight.

You want to have competitors who really put up a really great fight, and who have incredibly great ideas, and then you improve your ideas. It's like when somebody runs the four minute mile. Once people see that it can be done, then they will be able to do the same thing. And so the four minute mile run isn't copied by the next competitor, he achieves it through competition.

But every once in a while our competitors do completely crazy things and they collapse by their own craziness and due to lack of hardcore and disciplined technical evaluation of what they are doing. I mean, hype is one thing, but believing your own hype is unforgivable.

JB: What are you referring to?

SIMONYI: The NC, for example. I think that the people around the NC started with some valid concerns. There is a price concern, which is not that great, but it is there. Obviously, if something is $900 it's better than if it's $1200. Certainly there are valid concerns in terms of cost of ownership, the problem of administration, and the issue of updating of software in a network. The boot time is a valid concern. But these concerns can be solved within the existing framework relatively easily. I mean it's not rocket science, it's a matter of attention, it's a matter of time; they will be solved.

The NC attempts create a whole new paradigm, where the user will be faced with a whole new set of tradeoffs but where these problems are allegedly solved. Of course who knows, because it does not exist yet, but it's plausible that the start-up time will be solved, or if there's no local state, that the administration problem is solved. It's plausible. It's not a hundred percent, because even then there have to be multiple servers, and when you update something it has to go to multiple servers.

So it's not like there will be one server machine in the world and all you have to do is change that machine and all the networked computers are suddenly up-to-date. No. In some organization there will be 20 servers, and so the updates have to go to all the 20 servers, and so on so forth. And when you are talking about computers there is no real difference between 20 servers or 200 work stations, both of them involve communications, both of them involve synchronization, both of them involve data distribution yeah, one of them involves a few more cycles, but cycles are the cheapest things in the world, they are like dirt, they cost ten to the minus ten cents per. You can't pretend that the problems will be only solved by creating this new architecture, and that the other tradeoffs (things like you don't have privacy or flexibility to run the program you need or that you can't exchange media or can't take a diskette home or can't install the nice new voice card) will be accepted without a word by the customer.

And then there are the speeches by Scott McNealy, where he says that the office computers, by golly, really belong to the companies so they should be able to do whatever they want. This is strictly speaking true, but doesn't he see how irrelevant it is, or how annoying it might be to the person who's working with that computer. And I guess they could get into a shouting match, and the office worker would say well, in that case I'm not going to take any work home, or in that case why don't we have a punch clock and punch in and out, and lose all the flexibility and all the innovation that people have offered in the past. It's crazy to try to make such a radical investment on the basis of such dubious tradeoffs. I'm sorry, but the claimed benefits are perfectly obtainable within the existing Windows framework and will be available in the existing framework, at which point the NC companies will be left with nothing. Zero. Zip. Which will be very sad. And meanwhile they will have made a considerable investment. And then we'll be blamed for wiping them out or something.

THE REALITY CLUB

Alun Anderson, John Maddox, Lee Smolin

From: Alun Anderson
Submitted: 6.9.97

John,

I hope all is well with you — and thanks for confusing me with Philip Anderson (electronically at least). I've always wanted to be a grumpy genius.

Anyway, I've been meaning to drop you a line for quite a while to say that I really like what your doing with The Edge and your website. The piece with Brian Goodwin was particularly fun for me.

I saw him a week or so ago at the party after John Maynard Smith's Darwin Lecture at LSE. All the usual suspects were there including Lewis Wolpert, Helena Cronin, Steven Rose, Rupert Sheldrake, Oliver Morton, etc etc including Lee Smolin too. I took John Horgan along and he was savaged by everyone, especially Lewis who was just reviewing The End of Science for The Evening Standard. It was a bloody round two for Horgan as the evening before I chaired a debate between him and Lee Smolin plus Roger Penrose. John started off with a handful of supporters but lost them during the course of the debate — Penrose and Smolin are not guys for journalists to pick intellectual quarrels with. But it was fun and we pulled a big audience.

I hope you saw the big feature we did on Lee. It explains his ideas far more clearly than he can!

Keep up the good work. Sorry I missed you when you were in London. Are you coming over for the Rhone Poulenc prize? I've been senselessly busy as I'm being appointed a director of IPC Magazines as well as running New Scientist.

Best wishes

alun

ALUN ANDERSON is the editor of New Scientist.

From: John Maddox
Submitted: 6.9.97

Dear John,

First, many thanks for your kind invitation for this week-end. Sadly, I cannot make it. I have a raft of commitments starting at the week-end — talks and things like that.

There's some gossip to pass along. I found myself panning Horgan (for reasons that will nor surprise you) and Smolin, which may be more unexpected. My objections to his argument? As you can imagine, I'm going to welcome anybody who can dispense with the need for a single Big Bang. I have the following points:

1. Do the universes inside each black hole have the gravitational mass that shows up in our Universe: if so, the chance that they produce stars must be pretty small?

2. Presumably "our" Big Bang is the inside of some other universe's black hole: but you still need inflation to smooth it out, so that it's a much bigger event than you would think by looking in our light-cone.

3. Where do the small variations between universes come from, when a Big Bang is a Big Bang, producing the well-calculated ratio of hydrogen to helium?

4. I believe the natural selection analogy is false. (I should have chipped in on your earlier web-site argument, but I hadn't read the book then.) The essence of natural selection is the interaction between the changing environment and the genomes of evolving species. Here it's simply an optimisation problem: the universes that make the most black holes will become more common, because they reproduce each other more often.

My real worry is that Smolin leaps too lightly over the quantum gravity problem. He said at Hay that that will be solved "in a couple of years". Really? That's the old Hawking fallacy - that there'll be a Theory of Everything any day now, just because people recognise that it's become the central problem. I'd like to know what Ed Witten and his ilk think of the outlook.

I also worry that even the flags that Smolin waves about his idea being a "speculation" will not prevent the man in the street believing that the origin of life and of the Universe are now wrapped up.

Sorry to sound so sour.

All the best,

JOHN

JOHN MADDOX, who recently retired having served 23 years as the editor of Nature, is a trained physicist, who has served on a number of Royal Commissions on environmental pollution and genetic manipulation. His books include Revolution in Biology, The Doomsday Syndrome, Beyond the Energy Crisis, and the forthcoming What Remains to be Discovered: The Agenda for Science in the Next Century (The Free Press,US; Macmillan, UK).

From: Lee Smolin
Submitted: 6.18.97

First, I would like to say that I am very honored that a range of astronomers and biologists have found the proposals made in my book worthy of their thought and criticism. In looking for and developing the idea of cosmological natural selection I was motivated primarily to try to invent hypotheses about quantum gravity that would have testable consequences for observable phenomena. I was very happy to have found a way to do this, but I have been continually surprised that scientists who work in the realms that I've stolen from and trespassed onto have taken a positive interest.

Having said this, let me take up John Maddox's points, in reverse order.

First, am I somehow acting irresponsibly towards "the man in the street", by discussing speculative ideas in a book meant for a wide audience? This is a point he raised also in his review in The New Statesman. I must insist that I would not have done so were it not the case that the ideas lead to a theory that is testable. Having devoted three chapters (8-10) and an appendix to explaining how the theory is to be tested, as well as having listed and discussed every objection and counter-example known to me, I think I've done my duty. My claim is only that the idea is genuinely testable and refutable, and could easily be refuted in the near future, for example by the discovery of a 3 solar mass black hole. Given that many recent physical and cosmological theories that have been widely discussed in the recent popular literature are not so testable, I think I am doing nothing unethical in bringing the idea forward in this way, especially as I tell the reader many times which ideas in the book are accepted science and which are new, not yet confirmed, hypotheses.

Second, am I falling for the "old Hawking fallacy - that there'll be a Theory of Everything any day now"? I really don't think so. I've been working on quantum gravity for about 20 years and I've never before believed or said that we were within a few years of real progress. My belief that we are is based on several developments. There are in fact four directions along which progress has been and is being made; black hole thermodynamics, string theory, topological quantum field theory and non-perturbative quantum gravity. In each of these there have been dramatic developments in the last two years, capping many years of frustratingly slow (but steady) progress. In both string theory and non-perturbative quantum gravity we are now able to make specific and robust physical predictions about Planck scale phenomena. It is very tempting to believe that these four directions represent partial progress towards a single theory, in the same way that the developments of the first 15 years of the century in the thermodynamics of radiation, low temperature physics and atomic physics represented each partial progress towards quantum mechanics. Most importantly, in several different ways these different approaches are now being connected to each other. I personally believe that in the next years we will see a convergence of these four directions that will lead to a real quantum theory of gravity.

Of course, it is possible to disagree about this. We place our bets by investing our time and effort, and I am happy to leave it to the future to see how it turns out. But certainly it is true that in the last year a sense of excitement and optimism has been expressed by many people in these areas, especially string theory and topological quantum field theory.

4. Is the natural selection analogy false? As I explain in Chapter 7, there is a precise mathematical analogy between how cosmological natural selection works and the standard mathematical model of a species evolving by natural selection on a fixed fitness landscape. We can quibble about the meaning of words; I would agree that "the interaction between the changing environment and the genomes of evolving species" is crucial for the history of life and is not present in the cosmological theory I discuss. Whether or not this, or the optimization of numbers of surviving offspring, represents "the essence of natural selection" is a semantic question that in any case I don't have the authority to decide.

Actually, I think the interesting question is to what extent the standard mathematical model of a single species evolving on a fitness landscape is a reasonable representation of real biology. It is clear that it captures part of the truth, but it also leaves out the possibility of collective effects involving many species. Such effects are discussed by biologists over the whole spectrum from Dawkins to Margulis. Mathematical models by Bak, Sneppen, Paczuski, Kauffman and others seem to show that the existence of such effects is perfectly compatible with the basic postulates of neo-Darwinism.

3. This seems to be two questions:

"Where do the small variations between universes come from...?" The original point was to ignore this question, give our ignorance about Planck scale physics, and just to postulate that there is some source of variation. This is what Darwin had to do, as he was equally ignorant of molecular biology. However, given the progress in string theory, we can try to answer the question. Probably the best hypothesis is that it comes from transitions between different string vacua, of the sort discussed last year by Strominger and others.

"....when a Big Bang is a Big Bang, producing the well-calculated ratio of hydrogen to helium?" . As I explain in detail in Chapter 8 and in the appendix, the ratio of hydrogen to helium is under the control of several parameters, the most important one of which is the strength of the weak interaction. Assuming this varies like the other parameters, its present value, and hence the present ratio of hydrogen to helium, must be set in such a way that it maximizes the production of black holes. As I discuss in the appendix, (p 311.) this leads to a possible direction for testing the theory.

"2. Presumably "our" Big Bang is the inside of some other universe's black hole: but you still need inflation to smooth it out, so that it's a much bigger event than you would think by looking in our light-cone. "

I agree that inflation is likely needed to explain the flatness problem. The combination of inflation and cosmological natural selection leads to several more tests of the theory, as I discuss in the appendix (309-310.) It should be noted that inflation is about to be subject to rather severe test, from the data about inhomogeneities in the cosmic black body radiation that the Planck and Map satellites are expected to bring in. In particular we will know if Omega is really equal to one, as the simplest inflation models predict, or if it is less, as the astrophysical data so far seems to suggest. If Omega is less than one additional parameters must be finely tuned, which could be more work for cosmological natural selection, and lead to more tests of the theory.

"1. Do the universes inside each black hole have the gravitational mass that shows up in our Universe: if so, the chance that they produce stars must be pretty small?"

This is a standard question, I believe that Alan Guth has an appendix devoted to a careful answer to it. The essence is that as gravitational potential energy is negative, it can cancel positive contributions from matter, so that a universe with zero total energy (or mass) can be arbitrarily big and have arbitrarily many stars in it.

May I make a last remark? For me, the most important argument of the book is not connected with the specific proposal about universes "reproducing" through black holes. It is instead an argument I make in the last half of the book about the connection between theories of space and time and the questions of how the laws of nature were chosen so that our universe is full of structure and variety. The conclusion of this argument is that the relational point of view about space and time, which is realized in classical general relativity, needs to be joined to some mechanism of self-organization such as natural selection in order to complete the unification of general relativity with quantum theory. The reason, in brief (the argument is given in the last two Parts of the book), is that a relational theory of space and time, such as general relativity or any reasonable theory of quantum gravity, requires a complex world if space and time are to have their usual meanings. But, given the second law of thermodynamics, the laws that govern the physics of large scales must have certain characteristics if the world is to be permanently full of structure and complexity, and avoid equilibrium. Thus, the line of thought from Leibniz, Mach and Einstein must be joined to the discovery that there are physical processes that can produce structure and complexity, as we learn from Darwin and contemporary biologists, if it is to lead in the end to a coherent theory.

When taken together with the apparent lesson of recent developments in string theory, which is that the problem of unifying quantum theory with general relativity does not have a unique solution, this argument leads me to conclude that some historical processes analogous to natural selection must have played a role in our past to choose the laws of nature from a larger possible set, leading to a world with the structure and complexity of ours. I am not at all sure that the particular theory described in the book, involving black holes and "many universes" will survive. It is just the only theory like it I've been able to devise that was not easy to rule out by comparison with observation, while at the same time remaining genuinely vulnerable to falsification by plausible observations, such as three solar mass black holes.

Let me then end by coming back to the first point. What I think I have shown is that one way to wrest real falsifiable predictions from quantum gravity and string theory involves taking advantage of this last argument to posit that historical process analogous to those that built up structure in the biosphere have taken place on cosmological scales. If this is in fact right, then I believe it has huge implications for a whole range of philosophical issues from the methodology of the sciences to the foundations of mathematics to, perhaps, theology. Because I believe this is very possibly the direction that science will take, I thought that it was fitting to bring the argument to the attention of a wide audience of scientists, philosophers, intellectuals and people in general. It is because of this, and not to try to pull something over on "the man in the street" that I have tried to write for a wide audience, while at the same time trying to communicate honestly the real uncertainties and difficulties involved.-

LEE SMOLIN is a theoretical physicist; professor of physics and member of the Center for Gravitational Physics and Geometry at Pennsylvania State University; author of The Life of The Cosmos, forthcoming (Oxford.

Arnold Trehub & Steven Quartz on "Organs of Computation"

From: Steven Quartz
Submitted: 6.3.97

"If there can be innate circuitry and organization, there can be innate knowledge and skills - the two are simply descriptions of the same thing at different levels of analysis." — Steven Pinker

Pinker suggests that my list of constraints on the developing brain amount to innate knowledge, thereby reducing whatever genuine differences there might be between the positions. No differences-no problems. Well, not quite. I posited "initial cortical circuitry," not "innate circuitry." Although it might seem like splitting hairs, it is a key difference, one that when fleshed out reveals why Pinker's position is far removed from mainstream developmental neurobiology.

First, initial cortical circuitry (for sake of argument let's say the state of brain circuits at birth) can, and does, constrain without encoding domain-specific knowledge (as I consider below). Second, Pinker's theory of language as an instinct isn't just about whether the initial state provides constraints on development. It is a much more robust claim about how the mature state emerges. Pinker tries to minimize this difference by suggesting that much of development takes place prenatally, leaving little work for the postnatal period; he states:

"The plasticity has been described as some kind of "learning," but much of it goes on in the pitch-black womb, before sensory receptors have even formed, so it's better interpreted as part of differentiation and development."

This isn't the case for human brain development. Indeed, what makes human brain development so interesting is how slow and extensive its development is as it moves away from the sensory periphery. The real difference between these theories, then, concerns the processes underlying the generation of mature cortical organization, much of which operates postnatally.

Twenty or so years ago it was commonplace to encounter quotes like the following:

Linguistic information is innate in humans if and only if it is represented in the gene code (Fodor, Bever, & Garrett, 1974, p.450).

These kinds of claims were a long way from specifying any kind of neural mechanisms. One of my frustrations with much of the evolutionary psychology literature is that it does not go beyond this level of generality. Pinker's writing, however, goes beyond these generalities to be very explicit about what he's proposing; for example, he states:

"grammar genes would be stretches of DNA that code for proteins, or trigger the transcription of proteins, in certain times and places in the brain, that guide, attract, or glue neurons into networks that, in combination with the synaptic tuning that takes place during learning, are necessary to compute the solution to some grammatical problem (Pinker, The Language Instinct, 1994, p.322)."

This is innate circuitry-specified by genes that determine the pattern of cortical circuitry. We could imagine different sets of genes, each the product of natural selection, to encode the circuitry of each innate mental organ. In neurobiology, the closest advocate of such a position was Roger Sperry, whose chemoaffinity hypothesis once dominated developmental neurobiology. Actually, it is surprising that Chomsky never made reference to Sperry's theory. Sperry himself believed he had reduced developmental psychology to developmental biology, much as Chomsky argued would happen (i.e., that linguistics is ultimately a branch of biology).

I'm not sure how much work Pinker wants synaptic tuning to do. The traditional answer is not much. In fact, the hallmark of the tradition his position belongs to is that mature cortical structure unfolds through intrinsic maturational programs (grammar genes for language), in which the environment's role is exhausted by a simple "triggering" event. The environment is reduced to this role in part because it is too impoverished informationally to do any more.

The trouble is, Sperry's theory hasn't been viable for 25 years, and there are no contemporary neurobiological theories that could implement the sorts of developmental claims he and other evolutionary psychologists make. It is possible that Pinker will think I am creating a straw man, but I don't see any other candidate neural theories (though I'd be more than interested in hearing some candidates). There was a time when some hand-waving could be performed, citing how primitive our knowledge of developmental neurobiology is, but the rapid advancements of the field preclude this recourse.

What is so striking about contemporary developmental neurobiology (which was always thought to reduce to intrinsic programs) is how interactionist it is. This is the real and fundamental difference between nativist positions such as Pinker's and constructivist ones. The environment is not reduced to a minor role, but instead fundamentally shapes the developing cortex well into the second decade of life and indeed remains as lifespan plasticity. Among its defenders: Dale Purves, Larry Katz, Carla Shatz, Dennis O'Leary, Ed Callaway, and also selectionists: Edelman and Changeux. For this reason, I said that interactionist views are becoming the standard model in the field.

In his earlier work (a Cognition review of learning theory and some comments on Borer) Pinker speaks of the possibility that development is nonstationary, meaning that the mature brain organization (what used to be called the functional architecture) unfolds through interaction with the environment. Such a possibility was taken to be a major impediment to theory construction, and so was ruled out by methodological prescription, a principle Pinker called the continuity hypothesis. While this made developmental theory more tractable, it omitted what is most interesting about development-that the mature brain organization develops as a function of learning, becoming a qualitatively more powerful learning structure over time through its interaction with the world.

Such a possibility left an explanatory void-there was no vocabulary to describe this sort of developmental change. Learnability was stuck in inductive paradigms that were wholly inadequate to describe it. With the rise of computational neurobiology and theories of self-organization, however, a new vocabulary emerged, one that could begin to capture the richness of this interaction. This interaction is a form of learning, what Terry Sejnowski and I call constructive learning because it is a learning by building the brain. This is what makes the learning properties of developing cortex so intriguing, a type of learning that is much stronger than the learning straight-jacket traditional accounts shackled on the developing cortex.

In some ways this new perspective is even more radically different than what I intimated before. From its perspective, it is no longer even very interesting to be concerned about what must be innate. The old either/or dichotomies of rationalism/empiricism no longer generate the right questions. Arnold Trehub's interesting post was right concerning the need to understand how the initial state constrains development. And I think we are more or less in general agreement. But I would suggest that the generation of initial cortical circuitry does not reflect innate encodings of domain-specific knowledge. Instead, it constrains through more general properties, including a limited connectivity, that even in the human visual system is only rudimentary at birth. Because these constraints are not domain-specific, the developing cortex has a high degree of equipotentiality, a representational flexibility that is much better documented than Pinker's suggestion that they are friend-of-a-friend exaggerations.

From this perspective, the role of natural selection is also very different. But that's another story.

STEVEN R. QUARTZ, a fellow of the Sloan Center for Theoretical Neurobiology at the Salk Institute, has also been a member of the Computational Neurobiology Laboratory since 1988. He has advised the National Science Foundation on computational neurobiology-the use of parallel simulations to study development of the brain.

From: Arnold Trehub
Submitted: 6.11.97

Steven Quartz writes:

"Arnold Trehub's interesting post was right concerning how the
initial state constrains development. And I think we are more or
less in general agreement. [1] *But I would suggest that the
generation of initial cortical circuitry does not reflect innate
encodings of domain-specific knowledge*. [emphasis added]
Instead, it constrains through more general properties, including
a limited connectivity, that even in the human visual system is
only rudimentary at birth. [2] *Because these constraints
[initial cortical circuitry] are not domain-specific, the
developing cortex has a high degree of equipotentiality, a
representational flexibility* ......." [emphasis added]

If Quartz's claim [1] that there are *no* innate domain-specific cortical circuits were true then I can't see how normal postnatal development of the mind could occur. Nor have other investigators shown how it could occur. In my opinion, there are *some* minimal innate domain specific mechanisms which are essential for later adaptive restructuring of parts of the cognitive brain. I am led to this conclusion by the weight of empirical evidence, and by many years of work on formulating a biologically plausible neuronal model for human cognitive competence. In debating the issue, it is easy for each side to talk past the other. Perhaps a concrete example will help. There is one particularly striking phenomenon which demonstrates the likelihood of an innate domain-specific mechanism serving the perception of objects in 3-D visual space — the *moon illusion*.

We see the horizon moon as much larger than the zenith moon. This phenomenon has puzzled observers for at least 2500 years. It is now known to be an illusion because the true projected visual angle on the retina of either the horizon moon or the zenith moon is approximately 0.5 degree. Philosophers and scientists from Aristotle to contemporary psychologists have attempted and failed to provide a satisfactory explanation of the moon illusion. Some (e.g., Rock and Kaufman) have claimed that we judge the horizon moon as larger than the zenith moon because we compare the size of the horizon moon with known objects in our distant field of view, but are unable to do so with the zenith moon. It has been found that this explanation, which depends on experience-based learning of the relative size of objects, is wrong. In fact, there seems to be nothing in our developmental experience which can account for the systematic shrinking of the perceived size of the rising moon. A recent book (The Moon Illusion, 1989) reviewed the entire corpus of proposed explanations. In a summary chapter, the editor of the book, Maurice Hershenson, concluded that an understanding of the moon illusion was unlikely without an adequate theory of visual space perception.

In the course of my own work, I found that I was unable to account for *basic* visual-cognitive competence by any kind of post-natal learning process and I then accepted the necessity of an innate neuronal mechanism for representing objects in 3-D egocentric space, i.e., an innate domain-specific brain mechanism. The model that I proposed was motivated by the need to explain the neuronal basis of normal visual cognition, but it also happened to provide a neat explanation of the moon illusion as well as other visual illusions (see The Cognitive Brain, 1991, Chapter 14, "Illusions and Ambiguous Shapes: Epiphenomena of Brain Mechanisms").

A basic function of our visual system is to automatically maintain the perceived intrinsic size of an object despite changes in its projected retinal size as the distance between the object and the observer changes (size constancy). In The Cognitive Brain, I detail an innate neuronal system that maintains size constancy by automatically expanding or contracting the brain's representation of the changing retinal image in compensatory fashion as an object's egocentric distance increases or decreases. Briefly stated, the model predicts the moon illusion as a necessary consequence of a natural (evolution based) anisotropy in the size-constancy mechanism of our terrestrial-bound ancestors. The relevant anisotropy is determined by distance and vertical angle of egocentric representation. I think that the success of the model in accounting for the moon illusion is convincing evidence that the human brain is innately endowed with a mechanism for representing objects in 3-D egocentric space.

Quartz's assertion [2] about representational flexibility implies that initial domain-specific cortical mechanisms would preclude a high degree of neuronal equipotentiality. There is no empirical or theoretical basis for this view. It is important to recognize that the existence of innate domain-specific mechanisms in the brain is easily compatible with an abundance of remaining neuronal tissue having a "high degree of equipotentiality" and "representational flexibility". If we assume that, on average, a *new* neuronal representation is learned every two minutes during each 16-hour waking day over an 80-year period, I have estimated that this would require (allowing for a ten-fold redundancy) that only approximately 2% of roughly 13 billion cortical neurons be uncommitted and available for adaptive tuning. From this we can conclude that even if there were many innate domain-specialized mechanisms in the brain, it is unlikely that the remaining capacity for constructive neuronal development would be strained.

Steven Pinker and Steven Quartz (together with Terrence Sejnowski) have staked their opposing claims near the extremes of the nature-nurture continuum. In their forthcoming books, each will undoubtedly have some persuasive criticism of his opponent's position. But I wonder how soundly each of them can argue against the more credible view that *innate domain-specific brain mechanisms* and *learning-dependent brain structures* are *together indispensable* for the creation of the mature human mind. In this perspective, it cannot be claimed as a general proposition that one process is more important than the other. The respective roles of genetically-determined mechanisms (innate) and experience-determined structures (learned) can be teased apart and evaluated only with reference to *particular* cognitive processes and tasks.

ARNOLD TREHUB is adjunct professor of psychology, University of Massachusetts at Amherst, 1972 and the author of The Cognitive Brain, MIT Press, 1991