The human brain is a big, complicated system, with different parts doing different things. No one fully understands how it works, yet. But like many other researchers, I think I have a fairly good idea, at a high level.
...
Complete entry
Posted by
Pastor_Alex on 08/07 at 11:13 AM
Thanks for a nice summary of where we are with AI and neuroscience to this point.
Given that some researchers are suggesting that intelligence is not simply seated in the brain, but also anchored to the body in which we exist, do we not also have to understand the nature of that connection? Can we have intelligence without some physical manifestation or way of manipulating reality?
Posted by
b. on 08/07 at 01:13 PM
“It would mean that, once we’ve unraveled the specifics of how all the columns and their internals and interconnections work, then we could build a digital brain”
I think this is much more transhumanist ideology than fact, based on so many unproven assumptions that I don’t have time to deal with. I’ll just concentrate on the conflation of reality and simulation.
By definition if it’s “digital” it is not a brain, because brains are not digital. At best its a simulation.
Simulations are not reality because they are models because they are abstracted (details are removed to reduce complexity). All models have some degree of fidelity. How does one know what should be removed? Existing theory plays a role, but most of it is time and money. The more time and money the greater the fidelity. As far as I understand the spiking neurons models used in these simulations are still use Hodgkin-Huxley-like dynamics, which models the ion flow, but still is a very simplified model compared to actual neuron behaviour. A neuron is about as complex as a single cell organism. It still DNA, which effects and is effected by neuron behaviour. I know of no models of neurons that include DNA models, let alone all the neurotransmitters, neuropeptides, different types of neurons, glia, etc.
So we’re largely arbitrarily deciding on some level of fidelity/description and building models at that level, when we don’t understand how the whole system works, how can we know if we have removed the right stuff? This already assumes that there are components / processes in the brain that are not required, which seems implausible from an evolutionary perspective.
Models represent how we understand brains, and not necessarily how brains actually function. It’s an open question as to whether we can separate “actually functions” and “how we understand”. I don’t think they are separable. How we think about the world changes how we perceive it, and how we perceive the world changes how we think about it.
Posted by
R Wordsworth Holt on 08/08 at 11:48 PM
*All models have some degree of fidelity. How does one know what should be removed?
My feeling is that it is impossible to know in advance what can be removed. Include as much fidelity as possible within the constraints of time, money and computation. Theory driven trial and error may or may not yield interesting results or what might be considered success. At that point removing complexity can begin, again by trial and error - rather like a JENGA tower - keep going until it collapses.
IEET Blog