Paper code: 697
Title: The Enigma of John von Neumann's Theory of Automata
This is an ALife7 reviewer report on the above draft paper. The final, revised, paper, is available under the title John von Neumann and the Evolutionary Growth of Complexity: Looking Backwards, Looking Forwards... in the Alife7 proceedings, pp.~467--476. Essentially the same text is available online as technical report bmcm-2000-01 of the DCU Alife Lab.

A reassessment of von Neumann's achievement of self-replicating machines.

Quality of science:
      Importance of questions addressed: excellent
Significance of results: excellent
Soundness of methods: excellent
Quality of presentation:
      Clarity of logic: excellent
Overall readability and organization: excellent
Quality of English language: excellent
Relevance to conference:
Overall relevance to conference: excellent
Relevance to the theme "looking backwards": excellent
Relevance to the theme "looking forwards": average

Comments to Author (if any)
[Barry, I'm following your lead and foregoing anonymity. Mark Bedau,]

These comments are in no particular order.

1. The question of the distinction of trivial/non-trial self-replication is important in the field of ALife today, even aside from it's relevance to vonNeumann's particular concerns and intentions. This paper limits it's own scope to history: the correct understanding of von Neumann's achievement. Slight rewriting could bring out its larger significance to the field today--which would be good for the field.

2. The author should distinguish three questions: (1) Is JvN the *first* one to solve his problem? (2) Has JvN presented the *only* known solution to his problem. (3) Is JvN's solution to his problem the *simplest* known solution? He focusses only on question (3), but (1) and (2) are also important. In particular, the paper is written as if JvN's achievement would be fairly uninteresting if the answer to (3) were "No". But this rhetorical stance ignores question (1)--the answer to which is surely "Yes"--and this is enough to make JvN's achievement a landmark. So, I'd recommend some rhetorical rhevision.

3. I'd recommend a title change, to better reflect (and attract readers because of) the paper's thesis. Perhaps this: "John von Neumann solution to the problem of the evolutionary growth of complexity"

4. It would be useful to spend a paragraph comparing JvN's achievement with Conway's argument that the game of life will inevitably produce the evolutionary growth of complexity in an infinitely large field with random initial conditions (last page of his discussion of life in _Winning Ways_).

5. It would be instructive to add a paragraph on how *feasible* JvN's solution to the problem of the evolution of complexity is.

6. It would be instructive to add a paragraph discussing whether any existing ALife models (e.g., Tierra, known to be computational universal) also solve JvN's problem. In fact, this discussion would make the paper much more useful to the field. This comparison would highlight what is significant about JvN's solution, as well as clarify what the issue really is.

7. The author should be more explicit about whether the growth of complexity of interest is an *indefinitely* growing complexity--no ceiling. Of course, this issue is vague since the notion of complexity is vague. Still ....

8. This discussion seems to presume that the choice of "level of observation" is obvious or trivial--e.g., what to count as a "machine". But I think this is acutally a profound question. For example, if you focus on the simplest actual organisms (perhaps simple bacteria) and restrict your attention to that level, then you might miss the emergence of multi-cellular organizations. One strand of DNA looks a fair amount like another strand of DNA, even if only one of those strands is the code for a complex mutlicellular creature.

9. Could you clarify the following things: -- p. 5, left column, paragraph 3: the *class* of von Neumann's machines -- p. 5, r col, para 1: programmable constructor systems -- p. 6, r col, last para: how could the growth of complexity techniclly be achieved w/o self-replication -- p. 6, r col, last para: the difference between actual and potential growth of complexity, and why JvN's system won't show actual growth of complexity