Artificial and biological neurons communicate over the internet 2020 [pdf]
nature.com
nature.com
Can we change the post's title to the actual title of the article: Memristive synapses connect brain and silicon spiking neurons?
[0] I really dislike it when people try to punch up their brain-computer interface papers by adding an "internet" component to it. Unless it's somehow latency-sensitive, the internet part adds nothing over local control and is just pure clickbait. Please don't do this. If I review your paper, I will point this out, loudly, to the editor.
If you were going to build a brain-computer interface, why on earth would you scatter parts of it across three different countries like this? It's not exploiting a vast data center or anything, just a VLSI chip that happens to be far away. If you did have to do that, why wouldn't you directly connect all three centers instead of having a Python script(?) in the middle handing off information.
The membrane time constant of a biological neuron is also about 30 ms, which gives you a fair amount of wiggle room even in the face of lag.
Their solution seems clever, but other than a vague mention of "IoT brain implants" (please god, no), I don't see what problem it really solves.
Interesting. I had no idea it was that long. I figured it was on the order of 10s of ns.
Some very early work had estimated them as closer to 1 ms, but there were some technical problems with the measurement approach; I think that might account for some of the outliers there too.
Bear in mind that they're little bits of fat and water, so they're not going to be particularly fast. It's sort of amazing that the brain works at all!
[Also, no idea why your comment was dead, but vouched for it]
Memristive synapses connect brain and silicon spiking neurons
When I read the title I thought it was a metaphor for people using chatbots...
I'm more excited about adding more neuroplasticity, brainpower, and easier interfacing to humans, but if you can also make it so I can train NNs on rat brainware I'll be pretty happy, too. This further opens the path to both.
> Achieving this [communication over the Internet] is not trivial, since issues such as handling UDP propagation delays (which are typically variable and thus difficult to control) need to be resolved.
How would a backpropagation work in such an architecture?
Generally it is not known to my best knowledge how or whether the brain uses backpropagation (knowing that would be a huge result). However, several papers have been written on that topic with varying approaches. I encourage you to do a literature search if you want to dig deeper.
On this topic it is worth mentioning that it is not even known why the brain uses discrete signals (spikes) for communication.
In the simplest case you can feed signals 'in one end', let it do it's thing and then take the output, which you further feed through the rest of your architecture. Rewarding can be (in the simplest case) done with neurotransmitters (mostly dopamine) as it would've worked normally. You basically have to mimic something like this[0], by rewarding based on your loss function.
(of course, in reality, you'll have to do a lot more than that)