2 ms·

This kind of behavior in single celled organisms leads me to believe that there is a fair amount of complex logic handled by individual neurons in the brain, th

This kind of behavior in single celled organisms leads me to believe that there is a fair amount of complex logic handled by individual neurons in the brain, that is currently not being modeled in the state of the art neural networks A highschool biology textbook could have told you the same. Each neuron is an incredibly complex systems of pumps, channels, and pathways. There are many kinds of inhibition and excitation, both chemical and electrical. There's reuptake, active gene regulation causing dynamic protein synthesis, etc etc etc. Individual neurons can be multiple feet long. Every cell in your body is much, much more complex than we could ever imagine. our understanding of biological intelligence has a long way to go and correspondingly AI also has a long way to go before it can hope to simulate intelligence. That seems to assume mimicking the specifics of biology is necessary for progress. It's an open question, but that idea seems a bit silly to me.

mlyle 7y ago
The hard part is that we don't know the order of the problem. We know a neuron and synapse count for humans, but we don't know if that's minimal for intelligence. We don't know how complicated of a model of a neuron we need to get by with an equivalent number of neurons/synapses. Or what the tradeoff is-- how many more neurons we need for a simpler model, if that's even possible. Can we run smaller numbers of neurons to have a "slow" AGI? Does the fact that computers can operate the connections and weightings quicker earn us anything? And then you have the people like Penrose who argue there must be some quantum magic somehow being used by the neurons to perform quantum consciousness computation stuff. I mean, I suspect what you imply is right: we can get by with something much less than an exhaustive model of the human brain. But I can't make a rigorous argument for it, let alone how much less. P.S. There's a certain irony to be replying to someone named "solipsism" advancing this argument ;)

mlyle 7y ago

The hard part is that we don't know the order of the problem. We know a neuron and synapse count for humans, but we don't know if that's minimal for intelligence. We don't know how complicated of a model of a neuron we need to get by with an equivalent number of neurons/synapses. Or what the tradeoff is-- how many more neurons we need for a simpler model, if that's even possible. Can we run smaller numbers of neurons to have a "slow" AGI? Does the fact that computers can operate the connections and weightings quicker earn us anything? And then you have the people like Penrose who argue there must be some quantum magic somehow being used by the neurons to perform quantum consciousness computation stuff. I mean, I suspect what you imply is right: we can get by with something much less than an exhaustive model of the human brain. But I can't make a rigorous argument for it, let alone how much less. P.S. There's a certain irony to be replying to someone named "solipsism" advancing this argument ;)