[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]

2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

File: 111 KB, 695x583, rats.png [View same] [iqdb] [saucenao] [google]

3267562

Anonymous Wed Jun 22 14:11:56 2011 No.3267562 [Reply] [Original]

>A cortical neural prosthesis for restoring and enhancing memory

>A primary objective in developing a neural prosthesis is to replace neural circuitry in the brain that no longer functions appropriately.
>In this study, the application of a specially designed neural prosthesis using a multi-input/multi-output (MIMO) nonlinear model is demonstrated
>Also, MIMO model-derived patterns, delivered as electrical stimulation to the same electrodes, improved performance under normal testing conditions and, more importantly, were capable of recovering performance when delivered to animals with ensemble hippocampal activity compromised by pharmacologic blockade of synaptic transmission.

doi:10.1088/1741-2560/8/4/046017
You can get the paper for free if you sign up to the site.

Apparently, they've pretty much made a memory enhancing implant. Anyone with a CS or Math background care to go through their approach? All they've done is make rats pull levers, I wonder how much complexity their model could handle.

>>	Anonymous Wed Jun 22 14:13:56 2011 No.3267580 bump

>>	Anonymous Wed Jun 22 14:23:56 2011 No.3267621 bump1

>>	Anonymous Wed Jun 22 14:25:56 2011 No.3267633 How does it affect the choice? Looks like it just overwrites original brain's activity? If so It's more like making zombies with AI than enhancing anything.

>>	Axel the Possum Wed Jun 22 14:28:56 2011 No.3267647 >>3267633 From what I've read, it basically just simulates an artificial hippocampus, allowing long-term memories to be created and recovered. Not my field, though, so I could be totally wrong.

>>

Anonymous Wed Jun 22 14:39:56 2011 No.3267690

>>3267647
seems that's about right

>ensemble hippocampal activity compromised by pharmacologic blockade of synaptic transmission
means their neural prosthesis completely substitutes hippocampus

I don't know anything about brain anatomy though, wonder how do they make a connecting interface for such things. But it isn't something new as there already were experiments with rat's brain in a small robot body and i guess vise versa too

>>	Anonymous Wed Jun 22 14:48:56 2011 No.3267735 >>3267690 >how they make a connecting interface Just use an array of 12 electrodes or something like that. The nice thing about this is if it is small enough and can handle complexity you could think about selling it to people as a memory enhancer.

>>	Anonymous Wed Jun 22 14:50:56 2011 No.3267745 So what does it cost to get one of these babies installed?

>>	Anonymous Wed Jun 22 14:56:56 2011 No.3267779 >>3267745 Nothing. You can't. No doctor this side of the pacific will slice open your skull to hook you up to a roomfull of computers, just so you can remember how to pull a lever slightly longer.

>>	Anonymous Wed Jun 22 14:58:56 2011 No.3267789 >>3267779 do you know where I can get som ebooks on surgery then?

>>	Anonymous Wed Jun 22 14:59:56 2011 No.3267792 >>3267789 Ask /adv/ for an hero guides.

>>

Anonymous Wed Jun 22 15:00:56 2011 No.3267795

>>3267690
>means their neural prosthesis completely substitutes hippocampus
Well, no. It means it can mitigate the effects of impaired neurotransmission. The structure is still very much necessary.

Can anyone rapidshare the full article? I'd like to have a quick read.

>>	Anonymous Wed Jun 22 15:02:56 2011 No.3267802 >>3267795 Just register on iop. It's available for free.

>>	Anonymous Wed Jun 22 15:11:56 2011 No.3267834 >>3267802 reading it now, will let you know what I think. I don't have a background in CS or mathematics, but I do have a masters in neuroscience.

>>

Anonymous Wed Jun 22 15:12:56 2011 No.3267839

OP here, after some thought it seems that for the coefficients to be calculated, the rat has to learn first. They don't seem to have done memory transplants between rats, so I assume they tried and it didn't work. This chip sounds more like insurance against bad memory rather than a workaround.

>>	Anonymous Wed Jun 22 15:15:56 2011 No.3267850 >>3267839 >They don't seem to have done memory transplants between rats, so I assume they tried and it didn't work. Lol, no. It doesn't work like that. Baby steps is what makes us progress, not trying a fucking memory transplant.

>>	Anonymous Wed Jun 22 15:17:56 2011 No.3267863 >>3267850 It's a trivial matter of taking coefficients derived from one rat's model, and plugging them into another rat. Probably the coefficients end up being rat-specific, but then again, why would they?

>>

Axel the Possum Wed Jun 22 15:18:56 2011 No.3267867

>>3267839

>They don't seem to have done memory transplants between rats

Most likely because this is a very new technology. Once they've isolated the signals that form memories (which they appear to be very close to doing), then a memory transplant might be possible. In that case, the possibilities of this device become staggeringly large.

>>	Anonymous Wed Jun 22 15:20:56 2011 No.3267879 >>3267863 >Probably the coefficients end up being rat-specific, but then again, why would they? Because they're derived from the individual neural substrate. CA1 and CA3 are fairly large regions.

>>	Anonymous Wed Jun 22 15:21:56 2011 No.3267887 >>3267867 >Once they've isolated the signals that form memories Arg.. no it doesn't work like that. Those coefficients are specific to the rat in question. Why else do you think a training phase is necessary?

>>	Anonymous Wed Jun 22 15:42:56 2011 No.3267991 File: 28 KB, 512x367, hipp.jpg [View same] [iqdb] [saucenao] [google] >>3267987 Cock, forgot the image. Here it is.

>>

Anonymous Wed Jun 22 15:42:56 2011 No.3267987

So yea... what they did was make a device which bridges the gap between two regions of the hippocampus. The CA1 (see image: 2) and the CA3 region (see image: 1). They basically circumvent the schaffer collaterals which connect the two. They train a rat, and train a model based in input from CA3. The input is side specific, so a pattern classification algorithm models the right- and left specific patterns of neural firing. Rats which are subsequently tested on the task, and are being stimulated in CA1 according to the pattern which was modeled, did better on the task. Rats which received stimulation according to a pattern which was opposite of the current trial (so 'left' stimulation on a 'right' trial) did significantly worse. Furthermore, when transmission in the schaffer collaterals is disrubted by a pharmacological agent, the stimulation improved performance.

They effectively circumvented a monosynaptic pathway, which is, well, kind of neat. My criticism is that this does not necessarily have to do with memory per-sé. Stimulation may well simply facilitate perceptual decision making for instance. That, plus their figures are kind of shit.

>>	Anonymous Wed Jun 22 15:47:56 2011 No.3268007 >>3267987 Err... I just read over what I wrote. Does it even make sense?

>>	Anonymous Wed Jun 22 15:53:56 2011 No.3268035 >>3268007 Yeah, it does. Thanks Anon.

>>

Anonymous Wed Jun 22 15:58:56 2011 No.3268053

>>3267987

That's what I said, just from hearing the description of how it works on NPR. I mean its basically synthetic neuron firing. I think the guys when talking to the media tried to over dumb it down too much. Can't wait to see this stuff really start getting developed though. We're probably on the brink of entering a new age, the information age is pretty much done, we can access more information than we can use at once from the most standard of computers its time to upgrade the processing capabilities of the human being.

>>	Anonymous Wed Jun 22 16:06:56 2011 No.3268103 >>3268053 >upgrade the processing capabilities of the human being Don't count on that happening within our lifetime though. Progress like this is great, but not even remotely close to 'uploading memory' or anything like that.

>>	Anonymous Wed Jun 22 16:12:56 2011 No.3268136 Bumping this to drown the wave of transhumanism with actual science.

>>

Anonymous Wed Jun 22 16:20:56 2011 No.3268179

>>3268053
Well, strictly speaking, technically they have made a device which enhances memory. The model watches a rat learn a thing, and can restore the learned response even if the rat later loses the memory.

Obviously it is specific to the rat (although again, what exactly is the correlation between different rats' coefficient sets?) and not very fine, being compatible with only one kind of memory (based on their results). But that's a different matter.

>>

Anonymous Wed Jun 22 16:26:56 2011 No.3268220

>>3268179
> The model watches a rat learn a thing, and can restore the learned response even if the rat later loses the memory.
Not exactly. What they've done is stimulate memory <span class="math">retrieval[/spoiler] rather than eliciting the same cortical activation pattern directly. Rats did not loose the memories. Pharmacological disruption of the schaffer collaterals does not disrupt memory itself in the sense that it destroys previously formed synaptic weights. An alternative explanation of the efficiency of their device is that is promotes pathways formed via operant conditioning (which also involves the hippocampal complex), rather than autobiographical episodic memory, which is kind of they way it's presented in the media. Lastly, the effects are very nuanced, it's no clear dichotomy between 'rember' and 'don't remember' during incongruent modes of stimulation.

>>

Anonymous Wed Jun 22 16:33:56 2011 No.3268279

>>3268220
>An alternative explanation of the efficiency of their device is that is promotes pathways formed via operant conditioning
Also, this is supported by the fact that stimulation during conditions without pharmacological manipulation actually <span class="math">improves[/spoiler] performance. This suggests the stimulus-response association is in fact what is being simulated.

>>

Anonymous Wed Jun 22 16:37:56 2011 No.3268318

>>3268220
Well, unless I misunderstand:
1) Schaffer collaterals are critical for maintaining the learned response.
2) They have a drug which impairs schaffer collateral function.

Looks like using the drug drops memory performance quite a bit (Fig.4) and stimulation together with the drug reverses this quite a lot.

But honestly, I don't understand the training procedure. So first the rat sees one lever, and pushes it. Then they wait for the rat to poke the photocell, which takes 1-30 seconds, and then they show two levers. So to get water, the rat has to push the lever in the other direction this time? Why even show two levers? What is the rat supposed to be learning?

>>	Anonymous Wed Jun 22 16:38:52 2011 No.3268324 http://www.youtube.com/watch?v=gKwSDJ45HoI

>>

Anonymous Wed Jun 22 16:41:56 2011 No.3268344

>>3268318
>1) Schaffer collaterals are critical for maintaining the learned response.
True. But only because memory retrieval is disrubted, not the memory trace. After the effects are gone behavioral performance recovers.
>Why even show two levers? What is the rat supposed to be learning?
It's supposed to press the lever opposite of the one pressed in the probe trial. If it presses the correct lever, it receives sugerwater, if it presses the incorrect lever, it receives nothing. The rat is supposed to learn the oposite relation between probe and target trials.

>>	Anonymous Wed Jun 22 16:44:56 2011 No.3268378 >>3268344 Oh I see. So what is the time they keep measuring for their plots? The one that's 1-60 secs?

>>	Anonymous Wed Jun 22 16:47:56 2011 No.3268397 >>3268378 It's the time between probe and target trials.

>>	Anonymous Wed Jun 22 16:49:56 2011 No.3268411 >>3268397 Doesn't the mouse decide that time?

>>	Anonymous Wed Jun 22 16:51:56 2011 No.3268433 >>3268411 The nose poke is only to confirm the rat is still engaged in the task. Target trial doesn't start right after the mouse pokes its nose in the hole, but after a fixed delay.