yoginho

yoginho, 5 months ago

This manuscript follows up on https://link.springer.com/article/10.1007/s10539-019-9716-9, extending our earlier argument from the specific discipline of evolutionary-developmental biology #evodevo to a broad range of network models in neuroscience, cell & developmental biology, ecology, and economics. /1

yoginho, 5 months ago

We trace the origin of network models not only to graph theory but also lattice models (eg the Ising model) in condensed-matter physics in the 1920s, showing that their application to other disciplines follows a complex entangled pattern of model template transfers. /2

yoginho, 5 months ago

Model templates capture an intertwinement of mathematical structure, computational tools, and theoretical concepts that depict a general mechanism applicable to any field displaying a particular pattern of interactions.
https://academic.oup.com/monist/article-abstract/97/3/280/1168447 /3

yoginho, 5 months ago

Network models are successful model templates not because they provide accurate representations of the target systems to which they are applied, but because of their generic ontology & associated conceptual framework that can be used to model a wide range of different systems. /4

yoginho, 5 months ago

We provide extensive examples for our argument, starting with connectionist models (going back to Hopfield neural networks) in (cognitive) neuroscience, arguing that an excessive focus on network structure obscures important phenomena linked to the dynamics of the system. /5

yoginho, 5 months ago

We then criticize network approaches in cell & developmental biology, focusing on scale-free networks, network motifs, & Davidson's genomic regulatory networks. Again, an excessive focus on structure severely limits our outlook because structure does not determine function. /6

yoginho, 5 months ago

Somewhat ironically, in the context of cell & developmental biology, reverse-engineered networks based on a connectionist formalism may provide a compromise between abstraction of network structure & a focus on dynamical properties of regulatory systems. /7

yoginho, 5 months ago

Further examples come from ecology, where time-invariant models fail to capture dynamic changes in feeding habits during an organism's life cycle, or the non-discrete dynamics of ecological succession and extinction events. /8

yoginho, 5 months ago

Finally, we study the fascinating entanglement of Ising's original lattice model of ferromagnetism and Schelling's economic model of segregation in urban environments, showing that template transfer occurs in non-obvious ways, with unexpected feedback between disciplines. /9

yoginho, 5 months ago

We frame the entire complex of problems that is illustrated by these examples as a mismatch in the possibility spaces generated by static network models & differential-equation models based on dynamical systems theory. /10

yoginho, 5 months ago

Simply put, the possibility space of static network models is too restricted to capture many relevant phenomena in the domains of application of network models. The restrictions mainly stem from the intrinsic nature of the underlying model templates. /11

yoginho, 7 months ago

There are two argumentative tricks in the paper.

The first is to use a very broad definition of a "computing machine" as a generic input-output dynamical system, before restricting the definition to the more customary Turing machine.

The other is hidden in the peculiar assumptions that go into the claim that all finite physical systems are simulable by a universal quantum computer.

yoginho, 7 months ago

Basically, this claim only works if you can formalize the whole universe into some sort of state space (that can then be approximately subdivided into finite subsystems that can be simulated).

Also: simulation only works "with arbitrary precision" but that is not the same as "perfect simulation" as Deutsch defines the term. In practice, the difference is obviously unimportant, but if you want to make the claim that physics is (quantum) computation then you have to do better than that.

yoginho, 7 months ago

Finally, I love how he wiggles around the topic of falsifiability, indirectly admitting that the whole idea is of a rather metaphysical rather than physical nature.

What can I say? I'm not sold on this. Not at all.

What I get out of this paper is this: universal quantum computing is a damn good way of simulating physical processes (if it will ever be implemented in a tractable physical system).

Interesting stuff. But no convincing evidence that physics actually is computation.

yoginho, 7 months ago

Also interesting: Deutsch suggests Q-logic depth is non-decreasing, just like entropy, in this 1985 paper.

This strongly presages the arguments made in the assembly theory paper I recently discussed in this blog: http://johannesjaeger.eu/blog/assembly-theory-is-cool.

Really not so novel after all, that one...

yoginho, 7 months ago

But what all these papers reveal: there are huge philosophical problems with (pan)computationalism.

It is founded on dodgy equivalences & very peculiar interpretations of the relation between physics & computation.

I'll have much more to say about that soon...

yoginho, 7 months ago

@alexwild @eLife

I signed the petition calling this out. Not because I agree with Eisen's insensitive & stupid tweet, or leftist persecution paranoia. For me, this is really not about free speech or the Palestinian cause in particular. It's about an editor of a scientific journal making their employer look bad in public.

Still, to fire someone for one silly tweet, that's a staggering level of intolerance and thin-skinnedness.

yoginho, 7 months ago

Landauer's principle is really just about asking: what are the physical requirements if I want to calculate something using a physical device?

That does not make every physical process a form of computation, or information a physical thing.

The distinction we have to make here is philosophical: it begins with the recognition that "computation" is a model of an activity that a human being can do by rote. We can construct mechanical devices and make them perform such activities for us.

yoginho, 7 months ago

That's very different from the claim that physical processes or mechanisms are forms of computation. They are not. They lack the symbolic content of computation, which is instead imputed on them by us when we perform physical computations.

If you fail to make this distinction, you mistake your map for the territory.

Computation (& information) are ways to model physical processes. Though computation can be physically implemented (obviously), it does not follow that physics is computation.