Shaking things loose: whether you build things like brains #ai; fix them #neurology, #psychiatry; or want to understand them #neuroscience; what do you think are the biggest factors or unknowns holding us back?
@NicoleCRust One of the things I worry about & am excited about is complexity of systems. As we measure more about the animal, or in an ANN where we can measure everything, how do we understand it all together? If we have all this information, how do we make sense of it? Can we extract bite-sized principles? Is that enough?
@mishaahrens @NicoleCRust
Agree 100%! That's why I'm excited :). It would be very hard (maybe impossible?) to understand complex, heterogeneous systems without these measurements. I think we don't have most the methods we need to make use of this data to get "understanding" of how all the pieces fit together. I think one of the factors holding us back is the lack of these methods (I'm betting on combinations of human & machine intelligence). I also think ideas of how to measure how well these methods work (measurements of measurements of measurements...) are one of our biggest holes.
@kristinmbranson @mishaahrens @NicoleCRust Measurements of measurements… reminds me of our attempt at quantifying the risk that an edge in the #connectome is wrong, from incorrect reconstruction of the postsynaptic arbor, by the number of synaptic contacts divided by the size of the dendritic subtree spanning all synapses. Annotates each edge in the wiring diagram with a [0, 1] value, or score, of robustness to reconstruction errors.
@kristinmbranson @NicoleCRust I think that sometimes, measuring more things can make the problem simpler because you don't have to infer (or ignore or average over) hidden variables.