Inverse Turing - Modelling Human Brain with Universal Computer

Turing showed that all computer programs could be transformed to run on a very simple 'Universal Computer'.

One of the big questions this transformation allows is: Will the program Halt?
[The Halting Problem]
This, in the guise of a program to calculate the result, was used to prove some problems are not computable.

Turing also devised a unrelated idea - the Turing Test - that was at the heart of Artifiical Intelligence (AI) research for many years: If you can't see or hear what/whom you're conversing with, can you tell the difference between a real Human and a computer program?

Turning the 'Turing Test' around: Is the Human brain built from computing elements?

This isn't quite "Is it a computer?", there are two elements the structure and the componentry from which it's built.

If the Human brain is a computing device, and if its organisation and 'programming' can be discovered, then it can represented on an electronic computer. This is an unstated idea underlying the Turing Test.

The act of discovering the organisation and connections/programming could be impossible directly:
  • every human brain is unique in structure and organisation, though built from the same components (neurons) and performing the same tasks (seeing, hearing, recall, ...).
    This can be trivially shown by 'plasticity' - when a person suffers a traumatic brain injury (like a stroke or lump of steel through the head) their brain keeps functioning and often regains lost capability. Tasks normally performed in one area are taken up elsewhere.
  • Because of the microscopic nature of neurons and their axon connections and the sheer volume of them, it's unlikely a non-destructive 'read' of the whole brain will be possible.
But the idea is still very useful, even if we are unlikely to ever be able to clone our brains.

Exact computer models of brain functions/sub-systems can be built, examined and experimented with.
This is not useful in AI for recreating functions like speech recognition - electronic computers work so differently to neurons/axons that it doesn't yield insight or help for a 'reimplementation'.

For disciplines like Psychology, Sociology creating models of brain function and interactions takes study to a new level - it allows precise and repeatable initial conditions and impossible experiments to performed.

Most importantly, it removes 'hand-waving' explanations and forces exactitude and direct evaluation of theoretical models.

Computer animation tools, like MASSIVE, already implement 'autonomous agents' programmed with simple human behaviour/responses. This is a perfect platform to model and test Sociology models.

No comments: