The following is a quote taken from a book titled Kinematic Self-replicating Machines (Freitas R., Merkle R.,) (Landes Bioscience, 2004)
Another common objection is that for a machine to make a duplicate copy it must employ a description of itself. This description, being a part of the original machine, must itself be described and contained within the original machine, and so on, until it appears that we are forced into an infinite regress. A variant of this is the contention that a machine not possessing such a description of itself would have to use itself for a description, thus must have the means to perceive itself to obtain the description.
In this thread, I will be arguing that this objection is completely sound, and that having a copy of a "code" (or a string, "genetic code", tape, or "DNA strand", "Description", or whatever you want to call it) stored inside the replicator, is an absolutely necessary aspect of self-replication. It is not a sufficient aspect -- it is a necessary one.
The crux of my argument is that the transmission and bequeathing of a code to the offspring is the principle difference between organic reproduction and inorganic crystal growth. This Code Principle also underlies the argument that fire cannot be considered to be living, even though it "reproduces itself" in a dry forest.
We begin be imagining a cellular automata simulation on a grid, in two dimensions. This particular simulation adheres to two constraints:
(1) Volume preservation
(2) Kinematic realism.
"Volume preservation" means that the number of alive cells in the simulation must remain constant. These alive cells can change their internal state, but they cannot appear out of nowhere, nor disappear from the grid. "Kinematic realism" means that cells are forced to change the state of the grid only by moving to adjacent empty squares. That is, they cannot spontaneously teleport to a distant location, nor can they pass through each other like ghosts.
As far as the rest of the dynamics of this simulation proceeds, that is left completely open for the programmer to decide. The number of states and how those states affect the movement of the cells is left to whatever reasonably effective algorithm the programmer desires. But in all cases, the two constraints above must be adhered to.
Searching for self-replication
Imagine that some cells reach a certain state, which for brevity, we will refer to as "State8". In state8 the cell does not move in any direction; it is stationary. We desire that if cells reach state8, they remain in it for a long time, where "long time" means that they can begin moving again only under very rare, particular cases. (In a finite state automata context, state8 would be a special state of the cell that often goes back to state8, causing a cell to "get stuck" in it very often.). In the whirling chaos of the simulation, clusters of cells that are grouped into state8 stick out obviously to the unaided eye. They are stationary clumps against a background of flickering cells bouncing around.
We add an additional "rule" to the simulation that if two cells remain adjacent in state8 for many cycles (say approx 50 cycle) that they form a bond between each other, and are stuck in state8 for additional time (or forever). The 'organisms' in this simulation can now be referred to as bounded clusters of stationary cells -- akin to ball-and-stick molecules.
To find a replicator, we desire a certain cluster of bounded cells, along with a dynamics algorithm, such that the following happens: If this certain-shaped cluster is inserted into a random environment, it will increase the chances of a duplicate shape to form later in time. That is to say, given a random environment of random-stated cells moving in random directions, if we place this "catalyst" into that environment, it will give rise to a copy of itself at a different location. In other words, the catalyst will auto-catalyze itself from a random soup of constituent cells. This particular shape must have "statistically relevant" auto-catalysis. A random environment of cells will form clumps, but only undifferentiated random clumps. In many cases, the insertion of a bound state will simply be "dissolved" by the ensuing chaos from the outer moving cells. A replicator is achieved if it is peculiar in its capacity to fill the environment up with copies of itself who in turn make more copies in a runaway chain reaction.
towards Organic Reproduction
To anyone who has achieved such a feat, the reward is anti-climactic. If the grid is made very large, it becomes obvious that this simulation is not producing a "living" organism per se, but what has been simulated is in fact, crystal growth. Replication is taking place, yes, that cannot be denied, but these things are not alive. The catalyst is more like an ice crystal on glass trips off a chain reaction to form more ice all over the glass. "More like" is an understatement here. The simulation is literally simulating the process of crystal growth.
So how do we bridge this gap between crystal growth and something that is actually reproducing itself. In brutal honesty, a number of extraneous aspects will needed to be added this simulation (semi-permeable membranes, possibly metabolism, and other topics that exceed the scope of this article.) In any case, what is primarily lacking here is the transmission of a recipe/code/string/tape from the parent catalyst bound clump to the child catalyst bound clump of cells. The individuality of the clumps is given only by a strict list of particular cells that comprise them.
What I am claiming here is twofold:
(1) If the simulation had transmission of a code to offspring, it would no longer be simulating crystal growth.
(2) The addition of a transmitted code to offspring is not some far-flung idea that fell out of someone's wild imagination. It is precisely how living organisms in the real world proceed in reproduction.
(3) In response to Hoffstadter's nonsense (below) I will say that some extremely complex super futuristic robot that builds a copy of itself though a mirror. I would say that robot would require a code to run its complex software. By a pure and logical observation, that software would have to be transmitted to the offspring AT SOME POINT. I would love to hear someone correct my logic here, because I cannot see any reasonable dispute.
In the book itself, Freitas and Merkle place two stars next to the above quote. The two stars lead to a rather elaborate footnote on the same page containing quote. The footnote is some bizarre story given by Hoffstadter about some super futuristic artificial intelligence robot building a copy of itself on a moon. I simply cannot take such fictiony explanations as valid arguments or points of debate. I do not believe that such fanciful futuristic imaginings can help us discover coherent principles of self-replication. I really do not want to quote this story, but I will do so only for the benefit of the curious reader of this thread.
Now that the dust has cleared, we can declare a number of principles of self-replication that necessary for the type of autonomous replication we desire rather than just merely having crystals grow.
Catalytic Principle
When placed into an environment of mostly noise, the agent will increase the probability that an identical copy will emerge at a later time. These copies should do the same, and give rise to runaway replication in that environment.
Code Principle
The offspring must inherit a code from their parent version. The topic of this article.
Selection Eating Principle (not discussed in this article)
By a membrane, or other mechanism, the replicator must SELECT portions of its environment to take into its own space, and which to reject. This is precisely what the word "eating" means in an abstract context.
Metabolism Principle (not discussed in this article)
When, in the course of extracting energy from its environment, the replicator must act in such a way as to perpetuate this process indefinitely. That is, the use of energy from the environment must approach an end-to-end reaction that helps ensure its own energy in the future. ("Don't run out of stomach acid" and/or "do not hunt your prey to extinction".)
Here is the Hofstadter quote. It is silly speculation and I feel embarrassed quoting it. It's not worthy of being repeated. Proceed with caution.
Hofstadter,
Imagine that you wish to have a space-roving robot build a copy of itself out of raw materials that it encounters in its travels. Here is one way you could do it: make the robot symmetrical, like a human being. Also make the robot able to make a mirror-image copy of any structure that it encounters along its way. Finally, have the robot be programmed to scan the world constantly, the way a hawk scans the ground for rodents. The search image in the robot’s case is that of an object identical to its own left half. The robot need not be aware that its target is identical to its left half; the search can go on merrily for what seems to it to be merely a very complex and arbitrary structure. When, after scouring the universe for seventeen googolplex years, it finally comes across such a structure, then of course the robot activates itsmirror-image-production facility and creates a right half. The last step is to fasten the two halves together, and presto! A copy emerges. Easy as pie—provided you’re willing to wait seventeen googolplex years (give or take a few minutes) What we’d ideally like in a self-replicating robot is the ability to make itself literally from the ground up: let us say, for instance, to mine iron ore, to smelt it, to cast it in molds to make nuts and bolts and sheet metal and so on; and finally, to be able to assemble the small parts into larger and larger subunits until, miraculously, a replica is born out of truly raw materials. This was the spirit of the Von Neumann Challenge...this ‘self-replicating robot of the second kind’.
The fact of the matter is that we really don't have "basic sound understanding" of how to build a lifeform in a laboratory from scratch, starting from raw material. In addition to not having any "basic sound understanding" of how to engineer such an artifact, science is really not in possession of a collection of principles of abiogenesis.
This is not an anti-science rant. The reason science lacks these principles is because abiogenesis is not seen happening in nature, so we have not collected any data on it. We see lightning, earthquakes, stars exploding, and weather changing. We see those phenomena all the time, and so we can start to collect so much data on them that patterns emerge and scientific theories can be formed. Once the theories begin to dovetail with each other "principles" and "laws" emerge within the science. Abiogenesis is not even past stage 1 of this process. Abiogenesis is therefore, (at this time) a bunch of arm-waving and conjecture.
No, if a machine is a 3D printer, it may have other modules unnecessesary for the replication process.
It may have a fax module, sound module, etc, all these modules has nothing to do with the replication process, because the replication process can be limited to uploading a blueprint to the processing module.
