Thursday 19 June 2014

Some thoughts about Chomsky

Edmund Bolles is currently posting about Chomsky on Babel's Dawn - and this led to me finding the review of Chomsky's book The Science of Language by Christina Behme. This reminded me of the time, some 40 years ago when I first asked about the relevance of natural language to my research in "human friendly" computers - and on trying to relate what people actually did with my (admittedly limited) understanding of Chomskey I decided the area was not worth exploring. As a result I posted the following comment:
Your post has got me really excited, especially after reading Christiana Bohme’s review “A Potpourri of Chomskyan Science” of Chomsky’s “The Science of Language.” In the 1970’s I was involved in basic research into how one might build a “white box Computer” which would work well with people – in contrast with the “black box computers” we all take for granted. My problem was I was working in a strongly technology oriented environment with little relevant cross-disciplinary input. One aspect of my work involved two-way language communication between the white box and the human users. To get some ideas I turned to Chomsky, who seemed to be the acknowledged expert. I didn’t get very far as it was clear that what Chomsky was saying had very little relevance to what I was trying to do. I now understand why! It seems that I abandoned what were potentially good research ideas because I was fooled by a theoretician who had his head in the clouds and his feet a thousand feet off the ground.
I have also followed you link back to your earlier blog and the book by Benjamin Bergen “Louder than Words” and have put it on my ASAP reading list as it seems to mesh in well with my current research. Now that I am well and truly retired I have time to do the blue sky research spin-off from the long abandoned “white box computer” project.
In fact I hope in the next few weeks to get a detailed paper relating to the long abandoned project which shows that there are no significant differences in the way that human and animal brains store information and make decisions. No species evolves a brain which is bigger than it needs – and a major limiting factor is the cost of learning the knowledge the brain needs to “pay for its upkeep”. However there is a tipping point where, with the help of a simple proto-language, three things happen. The first is the “cost” of learning drops – so more information can be learn. The second is that the “storage” becomes more efficient – meaning that a brain of a given size can learn more. The third is that the first two make it possible to develop more sophisticated tools – and as the language is a tool it means that language can improve itself and become more efficient. Once this tipping point is reach there is a run-away development of “intelligence” involving a loop involving fast learning, more capacity to learn, and yet more powerful language – with no biological change in the brain being necessary.

No comments:

Post a Comment