Notes on Kauffman’s: ‘Investigations’

Kauffman: Investigations (my comments in [ ] )

Order in nature not due to natural selection alone, instead is spontaneous and self organized.

Investigations asks two questions:
1. Could there be a 4th Law of Thermodynamics for open systems - a law that governs biospheres.
2. Living things a autonomous agents able to at on their own behalf. What must a physical system be to be an autonomous agent?

[ If you view the idea of an autonomous agent as an illusion of directed action rather than pre-determined, un-precomputable and complicated deterministic interaction, then Kauffman’s problem with entropy goes away, and his idea of autonomous agents doesn’t pose any question - in other words the premise of the book goes away. ]

Kauffman does point out that there are arguments against a general law for open systems:

Open systems can be represented as computer programs, of which there are an infinite number. For most programs there is no shorter description [of the outcome?] than running the program, therefore no law that will predict behavior.
These porgrams run on universal turning machines which are models of all open systems, therefore there is no law for all possible non-equilibrium systems.
Kauffman suggests there might be a law just for self-constructing open systems.
His candidate law for such systems is that they ‘tend to maximize diversity’ - i.e. what they can do next. He calls this maximization of the adjacent possible.

Kauffman definition:

AUTONOMOUS AGENT: a self reproducing system capable of at least one thermodynamic work cycle.

[ If one compares the idea of open systems as something like a whirlpool, one sees that the material that they consist of is constantly replaced and replenished they are not like chairs or rocks, but are channels which act like turing machines, where the channel itself contains structure. Now if one considers the Shannon model of communication Source -> transmitter -> channel -> decoder -> recipient, you can see that an open system is a channel which can act as a transmitter or decoder itself. If you also look at shannon’s definitions, there seems to be no logical reason why the source cannot be the transmitter and the decoder the recipient.
In other words one could possibly have a Shannon model consisting only of turing machine channels, which are paths in some kind of phase space that correspond turing machines and to a subset of particular turing machines that self build.]

WORK CYCLES: Work constructs constrains, and constrains are needed for work to be done [ why does Kauffman need to say this - directed energy = work by definition? ] .

Maxwell’s demon shows the difference between measurements that can be used to extract work from non-equilibrium systems, and those that cannot. It seems that systems self arrange to extract work e.g. a windmill aligning to the wind.

Kauffman contends that statistical and information entropy miss some central issues !

Kauffman thinks that we lack a concept of propagating organization, rather than dis-organizing - i.e. entropy.

[ Kauffman is way off here. Firstly, there is no real problem with Maxwell’s demon anymore, if one accepts that thermodynamic/statistical entropy and logical (information) entropy are equivalent. Secondly, the windmill problem is again part of what I will refer to as the teleological illusion, the ghost in the machine illusion that makes ordinary things things look like they have a mind of their own. What we are seeing is the mechanism by which channels auto configure to maximize flow, even if the channel itself has structure. The windmill self configure to maximize entropy increase in its environment. All living things do this - or to use the passive (to avoid the teleological illusion), are configured by the environment to do this.
I would replace Kauffman’s candidate law with: there is a tendency to create complex channels that maximize overall entropy production. ]

Kauffman doubts that we can prestate the configuration space of a biosphere, in the way that you can prestate the positions and momenta of all the particles in a gas in a container. So how can you calculate macroscopic properties in the way that you get temperature and pressure for a gas?

[ surely that is not true - we can’t know the positions and momenta of particles in a gas which is why Boltzmann’s definition of entropy is based on statistical averages. ]

[ idea: what if we call the perceived structure in a system due to a series of DYNAMIC EQUILIBRIA, i.e stuff like a whirlpool appears when there is a situation like an overflowing bathtub - water flows in from the faucet and spills over the top - the overall shape of the contained water appears the same. A whirlpool is like this but in a spiral.
The fact that the structure of dynamic equilibria look like the structure of non self reproducing equilibria such as chairs and rocks means that the two have sometimes been confused.
The apparent structure of a machine is formed from a series of dynamic equilibria which when connected serve to maximize the rate of production of entropy. The fabric of the machine can be as complex as necessary as long as the information required to build the channel is less than flows through it during (its lifetime?, or each thermodynamic cycle? not sure what the correct value is here.
The seeming autonomy of an agent looking for food is merely a high entropy channel being created]

A bacterium swimming up a glucose gradient is used by Kauffman as an example of an autonomous agent.

In complex chemical reaction systems, self reproducing molecular systems form with high probability.

Size of biosphere: 100M species, humans 100K genes which encode 100K proteins. The number of possible proteins created by biosphere is possibly around 10 trillion. Current experiments randomly create around this number in a test tube.

Oster and Perelson: Shape space. Humans make approx 100M anti-body molecules. Their shape space has 7 dimensions - 3 for space and the extra ones for other physical properties of binding sites.
Newts have fewest anti-body molecules - around 10K. Oster and Perelson guessed that 10K covers 1/e of total shape space - approx 37%. Therefore each anti-body equals 0.37/10K. 100M is enough to saturate shape space so that all shapes are covered. In addition to shape space for binding, presumably there is a catalytic task space.
Kauffman says that because we can have 100’s of millions of proteins in the lab, we have a universal biological toolbox and so can start to examine life itself.
Life depends on autocatalysis. Autocatalysis = reproduction (self reproduction).

Reproduction in cells is carried out by protein enzymes.

Kauffman imagines there might be general laws about reproduction in proteins.
He imagines protein A and protein A’ & A”. A = A’+A”.
Protein A allows A’ & A” to join to create more A.
But what if you have A & B and A creates B from fragments of B and B creates A from fragments of A, then we have a self reproducing AB system. Kauffman wonders if you could could have a self reproducing system with hundreds of proteins.

[ clearly this is a sensible idea ]

Kauffman’s origin of life question:

Might life arise with collective auto-catalysis of proteins (or similar molecules) rather than DNA-style template reproduction?

Experimental results: As diversity of molecules in a reaction system increases, the system starts to spawn auto-catalytic reactions at a critical threshold.

[ the above is very interesting ]

Kauffman wonders if you have this plus a membrane also self.