In 'How to Create a Mind', Kurzweil states that Kasparov could simulate 100k board positions in parallel. Which was roughly same working knowledge as, say, Shakespeare with 100k word senses (http://books.google.com/books?id=FCcXiBPurdEC - search for 'board positions' and read next 3 pages).
Further, he suggests that each of the 100k may require 100 pattern recognizers in neo-cortex. That means, the brain needs to configure 10 million patterns in total - these are either newly observed, or new relations formed with existing patterns.
I'm sure its not linear, but if you divide 10M patterns by 10k hours, that means the neocortex needs to add an average of 1 pattern every 3.6s, or 4k patterns per every 4 hours (ie, limit for day). Thats about 400k neurons / day re-purposed for learning through deliberate practice to learn 40 board positions / day in chess.
__________________________
Continuing the conjecture, what we know about about spaced-repetition and mass-repetition might provide some nice bounds for upper limit on how these neurons can be exposed to learning material over 4-hr period to optimize their use as pattern recognizers.
Could make a nice illustrative simulation to speculatively address your question about limits of concentration / learning reservoir.
Further, he suggests that each of the 100k may require 100 pattern recognizers in neo-cortex. That means, the brain needs to configure 10 million patterns in total - these are either newly observed, or new relations formed with existing patterns.
I'm sure its not linear, but if you divide 10M patterns by 10k hours, that means the neocortex needs to add an average of 1 pattern every 3.6s, or 4k patterns per every 4 hours (ie, limit for day). Thats about 400k neurons / day re-purposed for learning through deliberate practice to learn 40 board positions / day in chess.
__________________________
Continuing the conjecture, what we know about about spaced-repetition and mass-repetition might provide some nice bounds for upper limit on how these neurons can be exposed to learning material over 4-hr period to optimize their use as pattern recognizers.
Could make a nice illustrative simulation to speculatively address your question about limits of concentration / learning reservoir.