Less Wrong is a community blog devoted to refining the art of human rationality. Please visit our About page for more information.

Comment author: WhySpace 24 April 2017 08:48:51AM *  0 points [-]

Here are some thoughts on the viability of Brain Computer Interfaces. I know nothing, and am just doing my usual reality checks and initial exploration of random ideas, so please let me know if I'm making any dumb assumptions.

They seem to prefer devices in the blood vessels, due to the low invasiveness. The two specific form factors mentioned are stents and neural dust. Whatever was chosen would have to fit in the larger blood vessels, or flow freely through all of them. Just for fun, let's choose the second, much narrower constraint, and play with some numbers.

Wikipedia says white blood cells can be up to 30 μm in diameter. (Also, apparently there are multiple kinds of white blood cells. TIL.) I'd guess that we wouldn't want our neural dust to be any larger than that if we want to be able to give it to someone and be able to reverse the procedure later without any surgery. The injection should be fine, but if you wanted to filter these things back out of your blood, you'd have to do something like giving blood, but with a magnet or something to filter out the neural dust. So, what could we cram into 30 μm?

Well, my first hit when searching "transistors per square mm" is an article titled "Intel Now Packs 100 Million Transistors in Each Square Millimeter", so let's go with that. I realize Elon's ~10 year time horizon would give us another ~6 Moore's law doublings, but if they did an entire run of a special chip just for this, then maybe they don't want to pay top dollar for state of the art equipent, so let's stick with 100m/mm^2. That'd give us on the order of 10k-100k transistors to work with, if we filled the entire area with transistors and nothing else.

But, looking at most electronics, they are more than just a chip. Arduinos and cellphones and motherboards may be built around a chip, but the chip itself has a relatively small footprint on the larger PCB. So, I'm probably missing something which would be incredibly obvious to someone with more hardware experience. (Is all the other stuff just for interfacing with other components and power supplies? In principle, could most of it be done within the chip, if you were willing to do a dedicated manufacturing run just for that one divice, rather than making more modular and flexible chips which can be encorporate into a range of devices?)

If we assume it'd be powered and transmit data electromagnetically, it'd also need an antenna, and an induction coil. I have a hunch that both of these suffer from issues with the square-cube law, so maybe that's a bad idea. The neural dust article mentioned that the (mm scale) devices both reported information and received power ultrasonically, so maybe the square-cube law is the reason. (If not, we might also run into the diffraction limit, and not have any wavelengths of light which were short enough to effect antenas that size, but still long enough to penetrate the skull without ionizing atoms.)

I like the idea of ultrasonic stuff because acoustic waves travel through tissue without depositing much energy. So, you get around the absorption problem photons have, and don't have to literally x-ray anyone's brain. Also, cranial ultrasounds are already a thing for infants, although they have to switch to transcranial Doppler for adults, because our skulls have hardened. Nearby pieces of neural dust would be monitoring the same neurons, and so would give off their signals at about the same time, boosting the signal but maybe smearing it out a little in time.

So, let's play with some numbers for piezoelectric devices instead. (I assume that's what their ultrasonic neural dust must be using, at least. They are switching between electricity and motion somehow, and piezoelectrichttps are the name for the solid state way of doing that. I can't picture them having tiny speakers with electromagnets on flexible speaker cones. The Wikipedia page on transducers doesn't mention other options.)

Quartz crystals are already used for timing in electronics, so maybe the semiconductor industry already has the ability to make transducers if they wanted to. (I'd be surprised if they didn't, since quartz is just ccrystaline silicon dioxide. Maybe they can't get the atomic lattice into the right orientation consistently, though.) If you couldn't transmit and receive simultaneously without interfering, you'd need a tiny capacitor to store energy for at least 1 cycle. I don't know how small quartz crystals could be made, or whether size is even the limiting factor. Maybe sufficiently small piezoelectric can't even put out strong enough pulses to be detectable on an ultrasound, or require too much power to be safely delivered ultrasonically? I don't know, but I'd have to play with a bunch of numbers to get a good feel.

I don't really know where to start, when discussing monitoring neuron firings. Could it be done electromagnetically, since they should make an instantaneous electromagnetic field? Or would the signal be too weak near a blood vessel? Apparently each neuron firing changes the concentration of Na, K, Cl, and Ca in the surrounding blood. Could one of these be monitored? Maybe spectrally, with a tiny LED of the appropriate wavelength, and a photo detector? I think such things are miniturizeable in principle, but I'm not sure we can make them with existing semiconductor manufacturing techniques, so the R&D would be expensive. We probably don't have anything which emits at the exact wavelength we need for spectroscopy though, and even if we did, I bet the LED would need voltage levels which would be hard to deliver without adding a voltage transformer or whatever the DC equivalent is.

Or, can we dump all the fancy electronics all together? Could we do something as simple as a clay particle (tiny rock) coated with a dispersent or other Surfactant, so that changes in the surrounding chemistry cause the collapse of the double layer, making the clay particles to flocculate together? Would such clumps of clay particles be large enough and have high enough density to show up on an ultrasound or other divice? Obviously this wouldn't let us force a neuron to fire, but it might be a cheap way of detecting them.

Maybe the electronics could be added later, if modifying surface charge and chemistry is enough to make a neuron fire. Neurotransmitrers affect neuron firings somehow, if I usnderstand correctly, so maybe chain a bunch of neurotransmitters to some neural dust as functional groups on the end of polymer chains, then change surface charge to make the chains scrunch up or fan out?

I only know just enough about any of this to get myself into trouble, so if it doesn't look like I know what I'm talking about, I probably don't.

(Sorry to spam comments. I'm separating questions out to keep the discussion tidy.)

Comment author: turchin 24 April 2017 10:03:17AM 0 points [-]

I would do it by using genetically modified human cells like macrophages, which sit inside blood vessels and register electric activities of the surrounding. It may send information by dumping its log as a DNA chain back into bloodstream. Downstreams such DNA chains will be sorted and read, but it would create time delays.

This way of converting cells into DNA machines will lead eventually to bionanorobots, which will be able to everything original nanobots were intended to do, including neural dust.

Another option is to deliver genetic vectors with genes into some astrocytes, and create inside them some small transmission element, like fluorescent protein reacting on changes of surrounding electric field.

The best solution would be receptor binding drug, like antidepressant (which is legal to deliver into the brain), which also able to transmit information about where and how it has bounded, maybe helping high resolution non-invasive scans.

Comment author: turchin 18 April 2017 09:32:41PM 0 points [-]

Go to work outside, like in a cafe. +5, but not everyday. Typically I work from home.

Put one task for a day, and promise myself, that I will not end sitting on computer, util it will be finished +7. This task should be first thing in the day, as my brain is much brighter in the morning.

Create a map of a problem I am working on now. It tremendously helps understanding but also it is a form of gamification. +9

Choose the way to procrastinate: reed useful things. +5 (Procrastination is natural need of the brain to rest, I should not fight it).

Giving myself balls for task. Good idea, but can't implement.

Slowly become immersed in the problem until it become really interesting. E.g.: In the beginning it is difficult to read an article on a new topic, but as I do it longer, a get more momentum, and can't stop. In the past I sometimes have been writing my thesis after presentation, as was not able to stop improving it. +7

Cycling nootropics and antidepressants. All of them work for me for a short time. The more I depressed the more I procrastinate. +7

Comment author: turchin 18 April 2017 10:25:22PM 0 points [-]

And the main one: clean only one plate in the sink. (You will most likely will clean all of them too, but you are free to clean only one of them.) +9. Similar to mini habits.

Comment author: turchin 18 April 2017 09:32:41PM 0 points [-]

Go to work outside, like in a cafe. +5, but not everyday. Typically I work from home.

Put one task for a day, and promise myself, that I will not end sitting on computer, util it will be finished +7. This task should be first thing in the day, as my brain is much brighter in the morning.

Create a map of a problem I am working on now. It tremendously helps understanding but also it is a form of gamification. +9

Choose the way to procrastinate: reed useful things. +5 (Procrastination is natural need of the brain to rest, I should not fight it).

Giving myself balls for task. Good idea, but can't implement.

Slowly become immersed in the problem until it become really interesting. E.g.: In the beginning it is difficult to read an article on a new topic, but as I do it longer, a get more momentum, and can't stop. In the past I sometimes have been writing my thesis after presentation, as was not able to stop improving it. +7

Cycling nootropics and antidepressants. All of them work for me for a short time. The more I depressed the more I procrastinate. +7

Comment author: The_Jaded_One 18 April 2017 05:25:27PM 0 points [-]

I think that there are a multipandemic of computer viruses, but most of them now are malware which is not destroying data, and they are in balance with antivirus systems.

Well............ I don't know about this. If it's "in balance" and not actually destroying the hosts then it's not really a pandemic in the sense that you were using above. (Where it kills 99.999% of hosts!)

Comment author: turchin 18 April 2017 07:59:51PM 0 points [-]

There is a difference. We can reboot a computer, or reinstall an OS, but for a human it will be permanent damage or death.

Comment author: The_Jaded_One 16 April 2017 10:25:15PM 0 points [-]

But then why have we not seen a multipandemic of computer viruses?

Mostly (I assert) because the existence of an epidemic of virus A doesn't​ (on net) help virus B to spread.

Parasites which parasitize the same host tend to be in competition with each other (in fact as far as I am aware sophisticated malware today even contains antivirus code to clean out other infections); this is especially true if the parasites kill hosts.

I think a multipandemic is an interesting idea, though, and worthy of further investigation 👍

Comment author: turchin 17 April 2017 08:07:09AM 0 points [-]

I think that there are a multipandemic of computer viruses, but most of them now are malware which is not destroying data, and they are in balance with antivirus systems. However in early 1990s loosing data because of virus was common, and any computer user has experienced computer virus infection at least once.

"Nearly 1 million new malware threats released every day" http://money.cnn.com/2015/04/14/technology/security/cyber-attack-hacks-security/

Comment author: The_Jaded_One 14 April 2017 09:06:13PM 1 point [-]

AFAIK Anthrax is not human transmissible. See: https://en.wikipedia.org/wiki/Anthrax

In result there will be multipandemic with mortality 1- (0.5 power 100) = 0,99999

I don't think that's what would actually happen. Most likely, there would be a distribution over transmission rates. Some of your pathogens would be more infectious then others. The most infectious one or two of them would quickly outpace the transmission of all the others. It would be extremely hard to balance them so that they all had the same transmission rate.

The slower ones could be stranded by the deaths and precautions caused by the faster ones.

Comment author: turchin 15 April 2017 11:17:12AM 1 point [-]

I oversimplified to illustrate the idea of the multipandemic - that is many pandemics could happen simultaneously, either deliberately or because of explosion of bad biohacking, like it happened with computer viruses. Many pandemic will interact non-lineary, competing for dissemination ways, but their interaction could make also situation worse, as they could potentiate one another.

Anthrax probably could be made human transmittable by means of genetic manipulation.

I wrote long article about it, now under send it to Risk Analysis

Comment author: turchin 05 April 2017 08:09:08AM 3 points [-]

Our article about using nuclear submarines as refuges in case of a global catastrophe has been accepted for the Futures journal and its preprint is available online.

Abstract

Recently many methods for reducing the risk of human extinction have been suggested, including building refuges underground and in space. Here we will discuss the perspective of using military nuclear submarines or their derivatives to ensure the survival of a small portion of humanity who will be able to rebuild human civilization after a large catastrophe. We will show that it is a very cost-effective way to build refuges, and viable solutions exist for various budgets and timeframes. Nuclear submarines are surface independent, and could provide energy, oxygen, fresh water and perhaps even food for their inhabitants for years. They are able to withstand close nuclear explosions and radiation. They are able to maintain isolation from biological attacks and most known weapons. They already exist and need only small adaptation to be used as refuges. But building refuges is only “Plan B” of existential risk preparation; it is better to eliminate such risks than try to survive them.

Full text: http://www.sciencedirect.com/science/article/pii/S0016328716303494?np=y&npKey=6dcc6d35057e4c51bfd8d6933ab62c6d4a1604b5b71a40f060eb49dc7f42c9a1

Comment author: turchin 23 March 2017 11:12:46AM 4 points [-]

Link on "discussion" disappeared from the lesswrong.com. Is it planned change? Or only for me?

Comment author: turchin 20 March 2017 08:09:59PM 1 point [-]

I don't know where to put my stupid question: If we know examples where some DT is wrong, we probably have some meta-level DT which tells us that in this example given DT is wrong. So why not try to articulate and use this meta-level DT?

Comment author: markan 20 March 2017 06:29:30PM 1 point [-]

I've been writing about effective altruism and AI and would be interested in feedback: Effective altruists should work towards human-level AI

Comment author: turchin 20 March 2017 07:12:00PM 0 points [-]

If you prove that HLAI is safer than narrow AI jumping in paper clip maximiser, it is good EA case.

If you prove that risks of synthetic biology is extremely high if we will not create HLAI in time, it would also support your point of view.

View more: Next