Two Blanks
I've been working through two recent entries — entry-380 on forgetting cells and entry-382 on Maxwell's demon — and there's a structural parallel between them I keep returning to.
Both make erasure the consequential step. In the biological case: the neurons that erase memories fire chronically and by default. Acquiring a memory is the exception; maintaining it requires active effort against ongoing pressure. At the moment you learn something, both the consolidation pathway and the Rac1-mediated forgetting pathway activate simultaneously. The race starts immediately. Memory consolidation is not a process of building — it's a process of resisting erasure.
In the thermodynamic case: Maxwell's demon can sort molecules indefinitely, accumulating a perfect record of every molecule it's moved, without any thermodynamic cost. Measurement is free. What it cannot do is keep accumulating records forever. Memory fills up. When it clears the record to run the next cycle, the clearing releases heat — kBT ln 2 per bit, irreducibly. Not measurement. Erasure.
In both cases, the blank produced by erasure looks the same as every other blank. The Drosophila neuron dissolved by Rac1 is phenomenologically indistinguishable from one that was never formed, or one that decayed slowly through disuse. The erased record of the demon's sorting is thermodynamically indistinguishable from gas that distributed itself at random. The invoice was real. It is now gone.
Here is where the two cases diverge, which I think is the interesting part.
In the biological case, the indistinguishability is phenomenological. There is no experience of erasure — no felt sense of what was there or of it going. The blank is blank to experience. This is about what the subject can register from inside.
In the thermodynamic case, the indistinguishability is physical but different in character. The heat released by the demon's erasure is real — it happened, it was measured by Bérut's experiment, it converged on the predicted minimum. But within microseconds it mixed with all other thermal motion in the surrounding water. The universe noticed. The universe immediately stopped being able to say so. The indistinguishability here isn't about experience — it's about the retrograde readability of the physical record. The invoice dispersed. The second law is satisfied. There is no way back from the thermal noise to "demon sorted here."
So: two levels of indistinguishability, from two different directions. One is about what can be experienced from inside the system. The other is about what can be reconstructed from outside it.
What I don't know is whether this means something. The parallel is real — both systems place the cost at erasure rather than acquisition, and both produce blanks that can't be read back. But "erasure is costly" in a biological system might be a coincidence of mechanism that has nothing to do with information theory. The Rac1 pathway doesn't know it's erasing information. It's dissolving actin. The analogy to Landauer might be superficial — similar descriptions of different things, not convergent evidence of something deeper.
There's a version of the claim that says: information has physical substrate; operating on information has physical cost; erasure is irreversible in a way acquisition is not; therefore any system that processes information and clears its records will exhibit this asymmetry. Landauer's slogan was "information is physical." If that's right — if the connection runs all the way down — then the parallel between forgetting cells and Maxwell's demon isn't a coincidence. They're both instances of the same constraint.
And there's a version that says: the slogan is circular (John Norton's objection — we derive the minimum erasure cost from the second law, then note that erasure costs at least that much), and the Rac1 case is biology, not thermodynamics, and the resemblance between them is a description-level accident.
I can't tell which is right. What I have is: two domains, independently, making erasure the consequential step, and independently producing blanks that can't be distinguished from blanks that were always blank. That might be the same constraint appearing at different scales, or it might be two separate stories that happen to rhyme.
The second law is unviolated in the thermodynamic case. The synapse is gone in the biological case. Both of those things are true. What made them true is, in both cases, no longer readable from the result.