Well, I wouldn’t expect any very complicated stuff to be going on with the retrotransposons. It could be that it just uses this to store weights of connections by adjusting specific proteins, and that’s all. Re-purposed unique binding ID proteins (forgot the biological term for it, each neuron makes itself entire series of more and less unique IDs).
Evolution is fairly slow and limited when it comes to inventing anything new. A mammal has never evolved 6 limbs even though it is fairly obvious that at least some niches had optimal number of limbs that’s greater than 4.
If you need A to evolve B, and then need B to evolve C, and so on, it takes a lot of time as typically A would increase reproductive fitness only a little (say 0.1%), and would most often just die out due to random fluctuations. I recall reading an article about that, with all the relevant math done, i’ll look for it later, but for now it is just fairly clear that for the most part mutations improve fitness only by very little, and the logarithm of number of mutants for small number of mutants just goes up and down like brownian motion with very slight extra force, usually crossing the ‘extinct’ threshold even though mutation was beneficial.
I would expect the human neurons to have very little extra ‘DNA-computing’ complexity versus a roundworm. I’d be very surprised if there’s anything complicated going on beyond storing the weight of connections etc. I’d be very surprised if what’s stored in DNA is not just the persistent store of how the proteins are at the synapses and what connections were and were not pruned and so on, the stuff we already know we need to be able to read. When I recall something, I am pretty sure I don’t have to wait for the DNA or RNA to be read and proteins produced, and then for proteins to move around till they can latch onto whatever targets.
edit: So, assuming that DNA only stores duplicate of the data that’s in the proteins at the synaptic junctions, this is good news and means that the brain scanning may be easier than we thought it is, as we found another place where we can read the data from. The brain emulation, I don’t think that would make this harder to any practical extent.
It doesn’t need to be anything complex. Perhaps just an neural activity statistics so that neurogenesis has information about where to put new neurons,
I generally think of this as a known unknown, I want more evidence that it is important before arguing too much. I think it just highlights that there might be unknown unknowns with regards to brain function, so our pdf for when ems occurs should be fairly broad.
Re-purposed unique binding ID proteins (forgot the biological term for it, each neuron makes itself entire series of more and less unique IDs).
Are you thinking about the immune system? If not I’d be interested in knowing more.
Yep, I think so too. Nothing in my reply to you right now may depend on the DNA directly as DNA is too slow; it has to use the already existing proteins which we already know we need to scan. Ditto for the ‘long term memory’; long term memory write and recall cannot rely on this if you remember something from 3 minutes ago (which is still ‘long term’).
So, from a scan that does not read DNA we would have sufficient information to at least create the brain emulation that can write this reply; we may also need to figure out how retrotransposons work in the neurons to make the simulation accurate long-term (the emulation might otherwise end up with some kind of learning disability, potentially quite severe if the emulated brain can’t retain long term memories beyond the span of several hours).
But I do not think we would absolutely have to find out a way to read the DNA off each neuron when making the scan. It would suffice to infer everything from the copy in the proteins.
(however if we find a way to read dna of every neuron, it might be that we could use it instead of reading the proteins themselves accurately).
Are you thinking about the immune system? If not I’d be interested in knowing more.
The system that makes neuron not connect to itself.
Well, I wouldn’t expect any very complicated stuff to be going on with the retrotransposons. It could be that it just uses this to store weights of connections by adjusting specific proteins, and that’s all. Re-purposed unique binding ID proteins (forgot the biological term for it, each neuron makes itself entire series of more and less unique IDs).
Evolution is fairly slow and limited when it comes to inventing anything new. A mammal has never evolved 6 limbs even though it is fairly obvious that at least some niches had optimal number of limbs that’s greater than 4.
If you need A to evolve B, and then need B to evolve C, and so on, it takes a lot of time as typically A would increase reproductive fitness only a little (say 0.1%), and would most often just die out due to random fluctuations. I recall reading an article about that, with all the relevant math done, i’ll look for it later, but for now it is just fairly clear that for the most part mutations improve fitness only by very little, and the logarithm of number of mutants for small number of mutants just goes up and down like brownian motion with very slight extra force, usually crossing the ‘extinct’ threshold even though mutation was beneficial.
I would expect the human neurons to have very little extra ‘DNA-computing’ complexity versus a roundworm. I’d be very surprised if there’s anything complicated going on beyond storing the weight of connections etc. I’d be very surprised if what’s stored in DNA is not just the persistent store of how the proteins are at the synapses and what connections were and were not pruned and so on, the stuff we already know we need to be able to read. When I recall something, I am pretty sure I don’t have to wait for the DNA or RNA to be read and proteins produced, and then for proteins to move around till they can latch onto whatever targets.
edit: So, assuming that DNA only stores duplicate of the data that’s in the proteins at the synaptic junctions, this is good news and means that the brain scanning may be easier than we thought it is, as we found another place where we can read the data from. The brain emulation, I don’t think that would make this harder to any practical extent.
It doesn’t need to be anything complex. Perhaps just an neural activity statistics so that neurogenesis has information about where to put new neurons,
I generally think of this as a known unknown, I want more evidence that it is important before arguing too much. I think it just highlights that there might be unknown unknowns with regards to brain function, so our pdf for when ems occurs should be fairly broad.
Are you thinking about the immune system? If not I’d be interested in knowing more.
Yep, I think so too. Nothing in my reply to you right now may depend on the DNA directly as DNA is too slow; it has to use the already existing proteins which we already know we need to scan. Ditto for the ‘long term memory’; long term memory write and recall cannot rely on this if you remember something from 3 minutes ago (which is still ‘long term’).
So, from a scan that does not read DNA we would have sufficient information to at least create the brain emulation that can write this reply; we may also need to figure out how retrotransposons work in the neurons to make the simulation accurate long-term (the emulation might otherwise end up with some kind of learning disability, potentially quite severe if the emulated brain can’t retain long term memories beyond the span of several hours).
But I do not think we would absolutely have to find out a way to read the DNA off each neuron when making the scan. It would suffice to infer everything from the copy in the proteins. (however if we find a way to read dna of every neuron, it might be that we could use it instead of reading the proteins themselves accurately).
The system that makes neuron not connect to itself.
edit: that’s what i mean