Implanted Optogenetic Light Switch Lets Scientists Flip Neurons On and Off

the_newsbeagle writes: Optogenetics is a fairly new (and fairly awesome) research tool for neuroscientists: By using light to jolt certain neurons into action, they can study how those neurons function in the mouse brain. But getting the light to those neurons has been difficult. Previous systems have required either fiber optic cables that tether the mouse to a computer, or heavy head-mounted receivers. Now Stanford's Ada Poon has invented a tiny and fully implantable system that wirelessly receives the signal to stimulate, and uses a micro-LED to activate the neurons. The device will let researchers study brain function while mice are running around, interacting socially, etc.

Comments

The Optogenetic Lights Witch

[smittyoneeach]

The Optogenetic Lights Witch
Couldn't avoid the oncoming ditch
Parked there in her beard
Due to Burma Shave feared
Her limerick was thick and rich.

I've long seen this as one of the

[Rei]

...key steps along the path that would allow seamless transition into becoming a digital mind.

Now, one needs a more advanced version that can (by pulse frequency / pulse patterns, light frequency, directionality, etc) communicate with many neurons at once at an individual level (the neurons having been "primed" to this behavior by means of selective photosensitive chemicals, inserted genes or nanostructures - their task would be roughly the complexity of a RFID chip, but would have to be done at incredibly small scales and in a manner that will diffuse into neurons). And it's not enough to be able to simply activate or deactivate neurons, you need to also be able to gather enough data (neurotransmitter levels, structure, etc) from them to be able to accurately feed a digital model of each neuron - that is, two way communication is required. And obviously a single implant wouldn't be enough (the optical signaling would quickly degrade into noise, even ignoring issues of optical attenuation - there's nearly 100 billion neurons in the human brain). There would have to be many such implants scattered all throughout the brain only monitoring / controlling their local area.

But once you have such a system, with enough bandwidth, and an external computing system with good enough neuron models and sufficient computing power, one could begin:

1) Pick an arbitrary neuron and start transmitting its data.
2) Begin simulating its' behavior based on that data.
3) Induce apoptosis in that neuron while feeding the result of the simulation into all of the neurons that it was in contact with.
4) Move onto anyone of the adjacent neurons and repeat steps 1-3.
5) Continue on until there are no neurons left and the entire brain is simulated.

Of course, we're still nowhere near either the hardware and software requirements of being able to pull off such a system. But if we had such a system, one could very slowly and gradually transition from a physical mind to a digital one, with there never being two separate consciousnesses (avoiding the moral issue of simply "copying" the mind into a digital form and then killing the version that was left behind).

Re:I've long seen this as one of the

[The Real Dr John]

As a biologist I can assure you this would be a form of slow suicide, not a transfer of your mind into a digital form. What the neuron is doing electrically at any one given time is not the full scope of what it is doing, or can do. Considering that neuroscientists don't know how long term memory is stored, you would not be transferring memories, as well as not transferring all functions of each neuron. It might make for a fun sci fi movie, but it wouldn't work. But you were probably kidding anyway.

Re:I've long seen this as one of the

[Rei]

What the neuron is doing electrically at any one given time is not the full scope of what it is doing, or can do.

Which would be a good point if I had written "only monitor electrical activity. But what I actually wrote was:

you need to also be able to gather enough data (neurotransmitter levels, structure, etc) from them to be able to accurately feed a digital model of each neuron

Something that I later described:

Of course, we're still nowhere near either the hardware and software requirements of being able to pull off such a system.

Furthermore:

It might make for a fun sci fi movie, but it wouldn't work.

Are you saying that given an arbitrary selection of data elements (implied by "etc" in the above - aka, "your choice as to what data to report"), and an arbitrary set of advancements in our knowledge of the behavior of neurons, that one will never be able to simulate a neuron? Are you saying that there are no physical measuable quantities that could allow neurons to be simulated? If so, you're moving into the realm of metaphysics, not science.

Re:I've long seen this as one of the

[Rei]

To put it another way: everything in the universe can be simulated.... given enough data and processing power. Of course, it's preposterously impractical to simulate an entire neuron from first principles, modeling every last subatomic particle, or to think that one could measure them all (even ignoring Heisenberg!). But of course, it's also preposterous to think that either of these things are requirements.

To pick a random example amount countless: What exactly would be the point to modeling a phospholipid membrane as countless individual molecules made up of even more numerous individual atoms made up of even more subatomic particles? It's just repeating phospholipid subunits. Of course it should be modeled in terms of its bulk properties, not individual particles.

But of course, does one even actually need to model the phospholipid membrane as some sort of physical object? Surely not - when looking at overall behavior, the particular details of the membrane's structure, apart from potentially a few adjustable parameters here and there, likely become statistically irrelevant compared to many reactions taking part in the neuron.

Modeling a neuron from physics first principles is technically possible (again, ignoring Heisenberg), even if practically impossible. But neither is anything close to that necessary. The neuron to simulate represents a system with inputs and outputs: X happens, Y results, the internal states Z are altered in manners A, B, and C. There exists some sort of mapping between the system's inputs, outputs, and states. The simpler one can encapsulate that into a model, the easier it is to simulate. The fewer the parameters that need to be measured, the easier it is to collect the data.

Re:I've long seen this as one of the

[The Real Dr John]

Simulating a neuron, and getting something to actually act exactly like a brain made up of billions of neurons are two completely different things. I am a biologist, not a hardware engineer, so I can only talk from the biology and neuroscience angles. I have had this discussion before here with people who are not biologists. Scientists still don't know how a single neuron works, not even close, so recreating it is still impossible. Plus, there are principles that we still do not understand, and therefore can not emulate them either.

Some people seem to think that if you could just imitate all of the electrical activity in the brain you would create a mind. That is silly. What do you think all those neurotransmitters and signaling agents are doing? The vast majority are not causing electrical activity in a postsynaptic neuron. They are causing biological (not electrical) changes in the surrounding membranes, and the intracellular structures associated with them. You mess with those chemicals, and you completely change the brain activity, both electrical and chemical.

So I am saying that in principle, what you are claiming is impossible. To make a brain, you are going to need to use wetware, not hardware. But of course you could prove me wrong by doing it, and I wish you good luck and success.

Re:I've long seen this as one of the

[Rei]

Simulating a neuron, and getting something to actually act exactly like a brain made up of billions of neurons are two completely different things.

Given the premise of an accurate simulation of a neuron and its interactions, how does that not logically imply the ability to simulate a collection of them by letting them interact in the same simulation, given sufficient processing power, memory, etc? If I can simulate the behavior of a single X, then I can simulate the behavior of a thousand Xs, a billion Xs, 1e100000 Xs, so long as I have sufficient computing hardware.

Scientists still don't know how a single neuron works, not even close, so recreating it is still impossible.

Which is the reason I wrote "Of course, we're still nowhere near either the hardware and software requirements of being able to pull off such a system". I'm well aware that we are not yet to the point of having a good understanding all of the properties of neural behavior to be able to make an accurate model. But I also think it unrealistic to expect that we will never be able to. There's nothing about a neuron that renders it forever immune to analysis.

Some people seem to think that if you could just imitate all of the electrical activity in the brain you would create a mind.

I am not those "some people", and if you had read my post instead of just assuming that about me, you would be aware of this point. I'm well aware that there is far more going on than just mimicking a pattern of synapses.

To make a brain, you are going to need to use wetware, not hardware

You have not presented a single argument to support this claim. Unless you think that neurons are physically knowable (evidence, please), that no amount of research can understand their working (wherein you're moving into metaphysics, not science), then they are simulatable (given sufficient research). And if one neuron is simulatable then an arbitrarily large grouping of them is, given sufficient computing resources.

Re:I've long seen this as one of the

[Rei]

Corr: That should have read "Unless you think that neurons are physically unknowable"

Re:I've long seen this as one of the

[The Real Dr John]

I look forward to seeing progress on this endeavor,but so far I have seen none. I can't wait for someone to prove my claim that hardware will not imitate a living brain wrong. By the way, most neuroscientists outside the field of cognitive neuroscience agree with me, and several have written entire books on the subject of wetware vs. hardware. Yes, there are neuroscientists who think it can be done, but they have not made any progress in that direction. Not even a little. Since you and I and everyone else do not know what "living" even means at the mechanistic level, I don't see how you can claim you can do the same thing with hardware. But good luck. I recommend you start by discovering the secret of life, and how it works.

Great to read about this

[ThatsNotPudding]

At least, that's what the lights told me to write.

Cool for research

[JoeMerchant]

In humans, an implantable optical "pulse generator" and fiber optic leads would be no problem at all to manufacture or implant (in a similar housing to modern pacemakers / neurostimulators.)

oblig Fletch

[sacrilicious]

Boyd: Oh by the way... what kind of a name IS Poon?

Fletch: Comanche Indian. Bye.

Using light ...

[PPH]

... to shut down neurons. They've just re-invented the television set.

Re:Maybe not that far away

[Rei]

Smallest rfid chip is .05mm x .05mm

Gee, I really wish I had written something like:

their task would be roughly the complexity of a RFID chip, but would have to be done at incredibly small scales and in a manner that will diffuse into neurons ... Of course, we're still nowhere near either the hardware and software requirements of being able to pull off such a system.

Instead of whatever it was that you think that I wrote.

No, I never talked about literally planting RFID chips. I said a task of roughly the complexity of a RFID chip. RFID chips are not the smallest possible unit (by many orders of magnitude) of what humans can build to handle a task of their complexity even in the present day. Today's top end CPUs are made with transistors of only 14nm (not um, nm) in size,. Neurons themselves do tasks many orders of magnitude more complex than an RFID chip, with their basic "design" coded into their DNA. Nanostructure self-assembly, which is increasingly important in technologies such as solar power and batteries, has the potential to reach RFID-level complexity. Etc. There are many routes one can take to produce that level of complexity at those scales.

Are we there yet? Hmm, what did I say about that? Oh yeah:

Of course, we're still nowhere near either the hardware and software requirements of being able to pull off such a system.