Anima Ex Machina

Originally uploaded by hzenilc.

Models and Simulations 2
11 – 13 October 2007
Tilburg University, The Netherlands

I attended this conference one month ago. Among several interesting talks, one in particular caught my attention. It was given by Michael Seevinck from the Institute for History and Foundations of Science at Utrecht, The Netherlands. His talk was about the foundations of Quantum Mechanics, and there were many NKS related topics that it brought  to mind. He talked about reconstructing Quantum Mechanics (QM) from scratch by exploring several restricted models in order to solve the so-called measurement problem, to deal with the nonlocality of quantum correlations, and with its alleged non-classicality, there being  no consensus on  the meaning of Quantum Mechanics  (Niels Bohr said once: “If you think you have understood quantum mechanics, then you have not understood quantum mechanics.”—More quotes of this sort on QM here).  The restrictons chosen in order to reconstruct the theory must be physical principles and not  theoretical assumptions. In other words, one approaches the problem contrariwise than is traditional, taking the least possible restrictions and exploring the theories that can be built thereon. The speaker characterized  this approach  as the “study [of]  a system from the outside” in order to ”reconstruct the model”. It is basically a pure NKS approach: “Start from a general class of possible models and try to constrain it using some physical principles so as to arrive at the model in question (in this case QM).”

One can then proceed to ask such questions as how one might identify QM uniquely, what it is that makes QM quantum, what set of axioms in the model is to be used, and which of them are necessary and sufficient? The question of meaning, previously asked of the formalism, is removed, and bears, if at all, only on the selection and justification of  first principles. Seevinck came up with the following interesting statement: “The partially ordered set of all questions in QM is isomorphic to the partially ordered set of all closed subspaces of a separable Hilbert space” (one of Mackey’s axioms in his axiomatisation of 1957). He added: “They (the principles)have solely an epistemic status. The personal motives for adopting certain first principles should be bracketed. One should be ontologically agnostic. The principles should be free of ontological commitment.” And further: “…axioms are neutral towards philosophical positions: they can be adopted by a realist, instrumentalist, or subjectivist.” He cited Clifton, Bub and Halverson who provided the following quantum information constraints used to derive quantum theory:

1. No superluminal information transfer via measurement.

2. No broadcasting

3. No secure bit commitment

Seevinck’s methodology in further detail is: Start with a general reconstruction model with a very weak formalism. Gradually see what (quantum) features are consequences of what added physical principles, and also see which features are connected and which features are a consequence of adding which principle. One thereby learns which principle is responsible for which element in the (quantum) theoretical structure.

One can generate further foundational questions over the whole space of restricted models, e.g.  how many of them:

- forbid superluminal signalling?

- allow nonlocality, and to what extent?

- solve NP-complete problems in polynomial time?

An important question which arises concerns whether intrinsic randomness would be of a different nature in different models or whether all of them would yield to deterministic randomness.

His talk slides are available online. Highly recommended.

Among other interesting people I met was Rafaela Hillebrand, of  the Institute for The Human Future at Oxford University. The Institute’s director, Nick Bostrom, has proposed an interesting theory concerning the likelihood that our reality is actually  a computer simulation. I have myself approached the  question in my work on experimental algorithmic complexity, in particular in my work on  the testability and the skepticism content of the simulation hypothesis. I will post on that subject later. The subject of thought experiments–in which I have an interest– was one that came up frequently.

Researchers at Berkeley working to unlock the potential of nanoscience:

High Definition Nanotechnology video from KQEDAmazing how nature produces its own nanodevices, such as motors like the flagella that allow spermatozoa to swim. Imagine how many structures can be found by exploring the universe of possible simple nanostructures! We also know that given a few elements, computing devices are capable of universal computation (see my previous post on the smallest universal Turing machine). So one could potentially provide  nanomachines with coded instructions to  perform just about any task–of course within the constraints of their mechanical capabilities.Further references available online from molecular to nano-computing:

- Tseng and Ellenbogen, Toward Nanocomputers, Science 9 November 2001.
- The world’s smallest computer made entirely of biological molecules, News Medica, 2004.
- Beckett and Jennings, Towards Nanocomputer Architecture
- DNA Computer Works in Human Cells, Scientific American 2007.

Alex Smith has recently been able to prove that a Turing machine conjectured to be capable of universal computation by Wolfram was actually universal (Wolfram 2,3 Turing machine Research Prize).

Part of the challenge was to find an encoding not doing by itself the universal computation that would make the Turing machine universal. Smith succeeded providing an encoding providing a background that while nonperiodic is sufficiently regular to be generated by infinite word written on the tape can be generated by a ?-automaton (not itself universal).

This relaxation of the tape content has been regular in the cellular automaton world, but also in the field of what is called weak-universality with the only difference that the Turing machines are allowed to start with other than blank tapes (blank tapes are periodic tapes with period 1).

An objection might be that such a coupled system could turn a nonuniversal machine into a universal one, but Smith showed that his encoding was capable of restarting the computation itself, something that coupled systems usually are not capable of. In other words, the preparation of the tape in Smith’s case is done in advance and the ?-automaton do not longer interact in any further time, while to have nonuniversal machines to become universal usually requires an automaton intervening at every step (or every certain steps) of a computation.

One may also be surprised by the need of an ?-automaton for infinite strings. But the difference to the traditional blank tape is subtle. Think of how machines operate in the real world. Machines do not run on blank tapes, they usually do so over memory (the cache, RAM or a HD) with all kind of information (that is considered garbage if it is not instantiated by a running program). You may think that this garbage is not infinite, but it is not so a blank tape for a Turing machine, so instead of thinking of providing a Turing machine with blank tape as need it, one can think of providing the Turing machine with a non-blank tape. Now, what is on the tape in the case of Smith is not exactly “garbage” because it plays a role in “helping” the Turing machine to perform its computation, in a kind of support on which the Turing machine that is capable of universal computation actually achieves the computation with the help of a defined background (that doesn’t mean that the machine cannot perform universal with other backgrounds) yet the background is not powerful enough to perform the hardest part of the computation. In the physical world, if processes are seen as computations, computations are performed on a support, on the background of the physical world. So these kinds of relaxation may be closer to actual physical situations than abstract situations in which a blank tape for a computation is assumed or required.

The discussion opened up by Wolfram’s work and motivated by Smith’s answer has generated a fruitful and renovated discussion of universality and complexity of small Turing machines. This is why I think this research s relevant to modern computer science, and not only as an amusing mathematical puzzle:

• New techniques for proving universality are found (Alex Smith’s novel approach for unbounded computations from arbitrary lengths and non-periodic initial configurations).
• New universal models of computation have been discovered (cyclic tag- systems, bi-tag systems).
• Such research provides a better understanding of universality,  its limits, its  underlying principles and its necessary and sufficient conditions.
• It is a base for actually building universal devices when only a few elements can be used, e.g. in nanotechnology or molecular computation.
• Simple/small machines may be more easily/effectively embedded in other systems.
• The old discovery/invention duality question comes to the fore: It sheds light on how simple universality is, how frequently it occurs, whether  it is engineered or not, whether  one builds universal computation or finds it in the universe.
• It could shed light on the relative feasibility of  universal Turing machines based on different tape configurations (e.g. blank characters, repetitive words, non-repetitive with computationally simple backgrounds) as actual physical systems.  At present it is not at all clear why one ought to  favor blank characters over other possible real-world backgrounds, such as “noise.”
• Questions of size and complexity  arise: It would be interesting, for instance, to find out whether there is a polynomial (or exponential) trade-off between program size and and the concept of simulating a process.
• Some questions  on algorithmic complexity arise: Will the encoding always be more complex if the machine is simpler? All theorems in algorithmic information theory depend on additive constants, which depend on the sizes of typical universal Turing machines. What is the impact of different generalizations of universality on algorithmic complexity and what is the role of  encoding in such a measure?
• Some questions arise on the relation between several variants of universality definitions: Is there an effective and efficient encoding for each non-periodic encoding preserving universality? If so, how does this impact their complexity? Is there a non-periodic encoding with blank characters for each periodic blank word encoding, and what would the impact of such  an encoding be on the size/complexity of the Turing machine in question?

The field is active and still an important area of research. Several computer science conferences include talks on small computational systems such as Computability in Europe (CiE) and Machines, Computations and Universality (MCU) included such talks this year, focusing in particular on reversible cellular automata and universal Turing machines.

Here are some references from the small Turing machine community:

[1] Manfred Kudlek. Small deterministic Turing machines. Theoretical Computer Science, 168(2):241-255, November 1996.
[2] Manfred Kudlek and Yurii Rogozhin. A universal Turing machine with 3 states and 9 symbols. In Werner Kuich, Grzegorz Rozenberg, and Arto Salomaa, editors, Developments in Language Theory (DLT) 2001, vol. 2295 of LNCS, pp. 311-318, Vienna, May 2002. Springer.
[3] Maurice Margenstern and Liudmila Pavlotskaya. On the optimal number of instructions for universality of Turing machines connected with a finite automaton. International Journal of Algebra and Computation, 13(2):133-202, April 2003.
[4] Claudio Baiocchi. Three small universal Turing machines. In Maurice Margenstern and Yurii Rogozhin, editors, Machines, Computations, and Universality (MCU), volume 2055 of LNCS, pp. 1-10, Chisinau Moldavia, May 2001. Springer.
[5] Turlough Neary and Damien Woods. Four small universal Turing machines. Machines, Computations, and Universality (MCU), volume 4664 of LNCS, pp. 242-254, Orleans, France, September 2007. Springer.
[6] Yurii Rogozhin. Small universal Turing machines. Theoretical Computer Science, 168(2):215-240, November 1996.
[7] Shigeru Watanabe. 5-symbol 8-state and 5-symbol 6-state universal Turing machines. Journal of the ACM, 8(4):476-483, October 1961.
[8] Shigeru Watanabe. 4-symbol 5-state universal Turing machines. Journal of Information Processing Society of Japan, 13(9):588-592, 1972.
[9] Stephen Wolfram. A New Kind of Science. Wolfram Media, 2002.

Stephen Wolfram’s NKS approach to the Reaction-Diffusion Process can be found at:

http://www.wolframscience.com/nksonline/page-1012g-text

A beautiful compound of different animals’  markings can be found in the following NKS book page:

http://www.wolframscience.com/nksonline/page-426