Tree F: wetware comparison. Summary in plain language
The question
Levin's argument began in living tissue. Companies now sell computers built from cultured human neurons, and these have learned to play Pong and Doom. If our ledger works, we can ask the substrate question in its cleanest form. Give neurons and silicon the same task, the same feedback and the same obstacles. Does either of them do more than we paid for?
What an experiment looks like
We agree a task with a partner who has a neuron-on-chip platform, and train a small software model to the same skill level under the same feedback. Both face the same obstacles. Then we put both through the accounting from Tree A and compare what is left over.
What the outcomes could be
- The neurons show a bigger unexplained residual. Anil Seth claims biology matters, and his claim would then have a number attached.
- Once training is matched, both show the same residual, and on this measure the substrate does not matter.
- The measurement cannot be made comparable across substrates. We would learn something about the ledger from that, and note it as a limit.
What kind of research this is
Collaborative and expensive. Everything depends on a partner and an ethics review. The tree is parked until the ledger exists and someone with a wet platform is interested.
Timeline
Year two at the earliest.