36 ms·

Say an old neuron dies. Could a 'replacement' neuron ensure continuity of the original 1000s of precise connections? Assuming that it can't, how is it to compe

Say an old neuron dies. Could a 'replacement' neuron ensure continuity of the original 1000s of precise connections? Assuming that it can't, how is it to compete with the other neurons? (There's a school of thought that neurons compete with other neurons for resources by making useful connections.)

gukov 8y ago
Just thinking out loud here, but is it acceptable to think that neural networks resemble UDP and not TCP?
- quantifica 8y ago
  Nah, bro. It's BGP, obviously. Packets, bro.
neom 8y ago
I would imagine if you could trigger something that required going there, brain might be able to allocate a spark that can re start the network provided no brain damage?
maxander 8y ago
Well, no, of course not- when a neuron dies, the information encoded in it's connections is lost. [1] But replicating the original connections isn't necessary, and certainly not necessary for the neuron to "compete." For example, say an old corporation dies, e.g. Microsoft [2]. Could a new corporation replicate the 1000s of precise products, services, and contracts of the original? Not at all. But replacement corporation could find new things to offer, potentially taking advantage of the economic shortfalls and new demands precipitated by the demise of Microsoft. Eventually the economy returns to full efficiency, though in a slightly different form than before. And just so with neurons- the information passed through the neuron that died will now be in demand by neurons downstream, and this demand can be fulfilled by new connections by the new neuron (or new connections from old neurons, or simply not.) But the information may come from different places, or have slightly different semantic value, or other differences. [3] Brains, like economies, change, and are good at changing gracefully. [1] At least, as far as we know- with glial cell participation and functional viral capsids and whatever else is going on, anything is possible, I guess. [2] Example chosen without malice. Honest! [3] All assuming the competition model of neural organization. Which is a plausible model last I checked, but see [1].
- fspeech 8y ago
  Except corporations don't really die that way. Even if the parent organization dies (goes bankrupt say), the children organizations normally survive with different ownership. So as long as a product or service has a market it won't really go away. A replacement would have to be in place first to kill the obsolete product or service.

gukov 8y ago

Just thinking out loud here, but is it acceptable to think that neural networks resemble UDP and not TCP?

quantifica 8y ago
Nah, bro. It's BGP, obviously. Packets, bro.

neom 8y ago

I would imagine if you could trigger something that required going there, brain might be able to allocate a spark that can re start the network provided no brain damage?

maxander 8y ago

Well, no, of course not- when a neuron dies, the information encoded in it's connections is lost. [1] But replicating the original connections isn't necessary, and certainly not necessary for the neuron to "compete." For example, say an old corporation dies, e.g. Microsoft [2]. Could a new corporation replicate the 1000s of precise products, services, and contracts of the original? Not at all. But replacement corporation could find new things to offer, potentially taking advantage of the economic shortfalls and new demands precipitated by the demise of Microsoft. Eventually the economy returns to full efficiency, though in a slightly different form than before. And just so with neurons- the information passed through the neuron that died will now be in demand by neurons downstream, and this demand can be fulfilled by new connections by the new neuron (or new connections from old neurons, or simply not.) But the information may come from different places, or have slightly different semantic value, or other differences. [3] Brains, like economies, change, and are good at changing gracefully. [1] At least, as far as we know- with glial cell participation and functional viral capsids and whatever else is going on, anything is possible, I guess. [2] Example chosen without malice. Honest! [3] All assuming the competition model of neural organization. Which is a plausible model last I checked, but see [1].

fspeech 8y ago
Except corporations don't really die that way. Even if the parent organization dies (goes bankrupt say), the children organizations normally survive with different ownership. So as long as a product or service has a market it won't really go away. A replacement would have to be in place first to kill the obsolete product or service.