Neurotwitter: A major AI stakeholder recently asked me if neuro understanding has advanced enough that we could simulate a human brain. I said no, but we do have confidence in relevant "known unknowns," and we mostly now have the tools needed to fill those knowledge gaps. Agree?
-
Show this thread
-
Replying to @neurowitz
Isn’t “simulating the brain” the wrong goal? To me it sounds like we’ll put in all the details and turn it on to see if it works, and if it doesn’t we’ll put more details in. Without a computational theory, simulating the brain looks hopeless to me.
1 reply 1 retweet 19 likes -
Replying to @dileeplearning
It's not a goal I would set. But it's a question that someone outside the field with billions of dollars to spend uses as a proxy for exploring how well we understand the brain.
1 reply 0 retweets 3 likes -
Replying to @neurowitz @dileeplearning
To address your question: How much experimental data is required to constrain/validate a computational theory of brain function? Surely it's a lot.
2 replies 0 retweets 0 likes -
Replying to @neurowitz @dileeplearning
Look at it this way: we can’t even understand and “simulate” the nervous system of C.elegans nematodes, and these worms have just 302 neurons, and a stereotypic and fully mapped nervous system.
1 reply 2 retweets 10 likes -
Replying to @OdedRechavi @dileeplearning
Do we know what's missing from our understanding of that system?
2 replies 0 retweets 1 like -
Replying to @neurowitz @dileeplearning
Every neuron has its own personality and tricks (see
@MHendr1cks studies for example), neurons communicate off the grid, and probably many other unknowns.@aexbrown could add for sure1 reply 0 retweets 8 likes -
Replying to @OdedRechavi @neurowitz and
I think the biggest gap is understanding/incorporating modulators and non-synaptic or non-canonical NT signaling. C. elegans as 300 (maybe 302) neurons but over 150 neuropeptides, many of which signal non-synpatically. The "network" they make does not match the connectome.
1 reply 9 retweets 31 likes -
Replying to @MHendr1cks @OdedRechavi and
Almost all mammalian neurons release 1 or more peptides, in addition to NT. The signaling properties are poorly understood, and they're rarely incorporated into models. We count spikes. Spikes are great, but a lot--maybe most--signaling between brain cells is not about spikes.
3 replies 5 retweets 33 likes
Yeah, and very dense and diverse in cortex:https://www.biorxiv.org/content/early/2019/01/13/519694 …
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.