[ExI] neurons

Anders Sandberg anders at aleph.se
Wed May 21 11:51:45 UTC 2014

Kaku seems to be talking about spiking and non-spiking states. Works for his purposes, since he is just making an order of magnitude argument that there is *way* more representational capacity in the brain than in the genome. A factor of ten doesn't change his conclusions much... in fact 4^3e9 is *way* less than 2^100e9=4^50e9 (and then to the Nth ~100 power for longer thoughts) and if there are more neuron states then the difference just becomes larger.
The issue of what states there are in the brain and how much information is encoded in spike patterns has been discussed endlessly in computational neuroscience - I slept through many lectures on it back in the late 90s :-) The gist of it is that rate codes are somewhat inefficient since you need to integrate a number of spikes to learn what the value is. Timing codes can be super-efficient *if* you can measure timing accurately and there is no noise... which there is. Distributed representations are likely and solve a lot of problems, but assigning information content to them is complex. Lots of information theory going on, no conclusion that really stuck with me except that the brain seems to use all methods simultaneously. 
But note that the change of rate in a rate code is a pretty slow and tricky way of representing something. We do seem to have neurons that turn such changes into signals encoding the direction of change instead, which are presumably easier to handle.

Anders Sandberg, Future of Humanity Institute Philosophy Faculty of Oxford University

William Flynn Wallace <foozler83 at gmail.com> , 20/5/2014 10:46 PM:
In The Future on the Mind, by Michio Kaku, he says as follows (facing page 342):

 "Define complex in terms of the total amount of information that can be stored.  The closet rival to the brain might be the info contained w/in our DNA.  Three billion base pairs containing one of four aids, therefore total amount of info is four to the three billionth power.  The brain can store much more - one hundred billion neurons, which can either fire or not fire.  Hence there are two raised to the one-hundred-billionth power initial states of the brain.... the states change every few milliseconds.  A simple thought may contain  one hundred generations of neural firings.  Hence there are two raised by one hundred billion, all raised to the hundredth power possible thoughts contained in one hundred generations.  Brains are ceaselessly computing.  Therefore the total number of thoughts possible within N generations is two to the one-hundred-billionth power, all raised to the Nth power.
My question concerns the underlined clause:  there are three states to a neuron:  increasing its rate, decreasing its rate, and staying the same.  Kaku says that a neuron fires or not.  This seems to say that a neuron is idle, waiting for stimuli, whereas I think that no neuron ever is not firing.
Am I confused again, or is he wrong?  bill w

extropy-chat mailing list 
extropy-chat at lists.extropy.org 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.extropy.org/pipermail/extropy-chat/attachments/20140521/db56543c/attachment.html>

More information about the extropy-chat mailing list