The Hal Baby Intuition Pump

Herb and Alice want to have a baby, but here’s how they do it:

1. They both have their genomes sequenced. They receive in the mail a data file consisting of their genomes, spelled out in two sequences of approximately three billion letters, A, C, G, T, . . . .

2. Then they write a little computer program that does the meiosis algorithm on both their genomes, (randomly) creating virtual sperm and eggs, which are then (randomly) united in silico, to create a new genome specification (which passes all the forensic tests for being the specification of the DNA of an offspring of Herb and Alice). (So far this is all done literally in terms of the symbols A, C, G, T, a purely computational process of string-rewriting.)

3. This specification is then used to construct, codon by codon, an actual DNA implementation of the entire genome, A = adenine, C = cytosine, G = guanine, and T = thymine. (Craig Venter’s lab can now do this.)

4. This genome is then implanted in the nucleus of a human egg (does it matter whose egg this is, since its own DNA will be removed before nucleus implantation?), and becomes a “test tube baby” in one of the usual ways.

Is the resulting infant, Hal, the child of Herb and Alice? It seems to me to be quite obvious that Hal is indeed their biological offspring, since it avails itself of all the genetic information they would contribute if Hal were conceived in the normal way. This intuition pump highlights the importance of what matters in reproduction: information, and the causal transmission of information (in this case, in the form of ASCII code for “A,” “C,” “G,” and “T,” not in the form of molecules). The causal link might, for instance, pass through telecommunication satellites, instead of taking the more direct, biochemical routes. [Drawn with revisions from personal correspondence, April 26, 2010]

Godfrey-Smith agreed with me that Hal is the offspring of Herb and Alice, but he had some reservations about my way of putting it (which can be read, unabridged, in Dennett, 2010; Godfrey-Smith, 2011). In the same spirit of constructive criticism, I acknowledged that there is a biologically important difference between the process Herb and Alice used to procreate and the way we normally do. What if everybody did it Herb and Alice’s way? Since the genetic information from Herb does not ride in its usual vehicle—a sperm cell—on its route to the egg, sperm motility would no longer be under selection pressure, and would, other things being equal, decline over the generations. Use it or lose it. Still, I maintain—and think this intuition pump illustrates clearly—that the molecular structure that stays roughly constant over the generations is preserved because it embodies the information that it does.

If there were alternate structures that preserved the information, evolution would continue unimpeded. This claim can be further examined in yet another intuition pump. Imagine that on another planet, “odd-numbered” generations used the DNA rungs A, C, G, and T, and “even-numbered” generations used a different double helix, call it XNA, with rungs made of P, Q, R, and S (some other molecules). We can suppose that the offspring’s XNA molecules were made from DNA templates from their parents by a mechanism rather like messenger RNA, but “translating” between the two different biochemical languages. These messages would be translated back by another messenger mechanism in the next generation, and so forth. You’d have to find a mate whose genes were written in the same language if you wanted offspring, but those offspring would have genomes written in the other language. Oedipal unions would be infertile—which is probably just as well—but there might be many Romeo-and-Juliet–style tragedies in which lovers could not procreate since they came from opposite communities. (They could always settle for a barren sex life and adopt babies of either type, or even play egg donor and sperm donor and raise half-siblings galore.) In such a world, aside from these complications, evolution would continue as always, transmitting genetic information about valuable adaptations (and heritable diseases, etc.) through the generations in alternative coding systems that could be as structurally different as you please. Same genes, different molecules. Each gene would have two forms, as different as “cat” and “chat,” “house” and “maison.” (Note the parallel with the case of the two black boxes: what the two syntactically, structurally different vehicles in both cases have in common is the same information, the same semantics.)