For those of you who don't already understand the Hall Effect, or its potential computational value in neurons, I will attempt to explain this WITHOUT references to obscure pysics or math you probably don't know.
When ANY charged particle travels through a magnetic field, its gets pushed sideways - the direction being dependent on the direction of the magnetic field. You might remember the coils that were on the necks of picture tubes that steered the electron beams. When this occurs within a thin conductor only a few molecules thick (I am thinking ion channels) the charged particles are pushed to the sides of the conductors. With really HIGH intensity fields, like electrical transmission lines, the forces pushing the electrons to the surface are so great that there is little current flowing in the centers of the conductors. In classical Hall Effect devices, the charged particles, usually electrons, and all pushed to a high-resistance surface so they "shut off" in the presence of an adequate magnetic field. Whatever happens at the sides of the conductors is entirely dependent on the nature of what is there. For example, instead of a high resistance surface as in classical Hall Effect devices, there could be substances that interact with the ions. The ions will have energy from the difference in electrical potential between the ends of the channels, and the substances on the walls of the ion channels might, for example, require some energy threshold to overcome to react. I suspect that a CLOSE look at ion channels will disclose more ways that ions could be affected by being pushed to the sides of ion channels than I can presently imagine. This could clearly create all sorts of non-linearities, delays, etc. Further, there is NO REASON to expect that parallel ion channels are working identically. Working together, several different ion channels could create really complex functions, perhaps without limit. The REALLY interesting part of this is that such closely spaced channels would have a mutually inhibitory effect from their respective magnetic fields pushing ions to the walls of adjacent channels. Further, groups of them would have their operation inhibited by the ambient magnetic field from surrounding phenomena like other neurons. Reciprocal inhibition is a well-known phenomena that everyone has presumed utilized MANY inhibitory synapses, when all that may be needed is for things to be near to each other so their magnetic fields restrict the flow of each others ions. A fellow at IBM a few years ago showed that ANY non-linear function of multiple variables could be implemented by running each variable through suitable non-linear functions, adding the results, and then running he sum through another suitable non-linear function. One simple example is multiplying numbers together by converting to logarithms, adding the logarithms, and converting the sum to its antilogarithm. Long ago I published a paper where, presuming for a moment that neurons were communicating the logarithms of probabilities of assertions being true, I computed what an inhibitory transfer function would have to be. Then, I looked at the carefully plotted transfer functions of the two (yes, just two) inhibitory synapses that have EVER been characterized (over the course of two years of experimenting). One was EXACTLY the predicted function, and the other was something COMPLETELY different. Neither was anything close to linear. More recently I amended this, showing that most likely things were computing with rates of changes in logarithms, rather than the raw logarithms of probabilities of various assertions. Making this last assumption greatly facilitates learning cause-and-effect phenomena. The bottom line here, given the geometry of ion channels, I see NO way that the Hall Effect could possibly avoid being a dominant force. OK, then there are the skeptics who have pointed out that we can hold a powerful magnet to our heads and be OK. However, the magnetic field a few molecules away from an ion channel where the next channel is located is probably MUCH more powerful than the GRADIENT in the field, an inch away from a rare earth magnet. Also, magnets are well known for their pain-reducing effects, which stupid "scientists" have dismissed because they couldn't see how magnets could do anything to something they saw as being fundamentally electrical and NOT magnetic. Hall Effects would explain this elegantly. So, how does this relate to AGI? It strongly suggests that we are WAY more complicated than previously thought, but that the "computational elements" (ion channels) may be simpler than whole neurons, or even synapses. All in all, simulation aside, if things work the way they appear (in light of this discussion) to work, then AGI's failure can be explained in part because it is two more orders of magnitude too simplistic to do interesting things. Thoughts? Steve ------------------------------------------- AGI Archives: https://www.listbox.com/member/archive/303/=now RSS Feed: https://www.listbox.com/member/archive/rss/303/21088071-f452e424 Modify Your Subscription: https://www.listbox.com/member/?member_id=21088071&id_secret=21088071-58d57657 Powered by Listbox: http://www.listbox.com
