Sunday, 15 January 2017

motherboard or decision-tree

Applying the most advanced and universally accepted principles of neuro-science (arising out of the hubris that computational powers could begin to map out every possible neural daisy and we’d soon understand how the brain works) to a system that humans (as inventors) understand better by degrees, the microprocessor yielded some fruitfully disappointing results.
The failure of the model of a sophisticated neural network (not a neural network itself but the parameters by which one is made) to understand arcade games—despite the demonstrations that machine-learning was able to give with little to no supervision—illustrates, I think, that despite the mechanical and philosophical differences between brains and circuits perhaps we still don’t have the framework and the context to glean meaningful, correlated results. What do you think?  Perhaps we cannot analyse the system we are in with the quiver of tools arising from the same.