I find this an amusingly cognitive view. I think the commentators do a pretty good job of dealing with the main issues. Yes, you could frame everything as 'prediction' but basically all you are doing is conflating 'prediction' with 'agency with a time delay.' This really doesn't solve any problems. At what timescale you want to draw the line between prediction and agency for closed loop systems is an open question. I think one of the key concepts that the author fails to address is that the fundamental problem the nervous system addresses is coordination of otherwise isolated muscles, organs etc. While we may remain mystified about the function of the mind, we do know that at some level the nervous system is really about control and synchronization. There are astoundingly good models (now with biological correlates!) of motor control loops that are just classic PID controllers. While the predictive perspective can be useful in some cases it is just a perspective and an interpretation. If you want to get to the heart of the matter you need math, and this paper is lacking on the math. To paraphrase: words are wind, math is truth.
I've never had that experience. It usually takes a lot more effort effort to work out what the math is actually describing and why the operations done on it make sense. That's if I even understand the symbols and functions they use. If not then it's totally inaccessible.
To be fair you can easily obscure natural language in similar ways.
Re math -- this is largely a work of interpretation, how to understand various results and place them in a common framework. Much of the work he cites is full of math and would probably be much more appealing to you, but he's not trying to build a new model. There's still plenty of philosophical groundwork to be done in the cognitive sciences.
Perhaps the brain just provides a place for the patterns to live? Perhaps it is not so much searching for patterns of activity which allow it to predict the future, but allowing the patterns of activity induced by the world to grow and evolve together. Then all it has to do is modulate this evolution when the organism gets something it likes. Reward modulated hebbian learning I believe it is called. I've not read the whole paper, but it seems overly complicated and unnecessary when you consider the above conception...
9 comments
[ 3.4 ms ] story [ 30.2 ms ] threadTo be fair you can easily obscure natural language in similar ways.