Brain Backups: What's My Brain Got to Do With Me?
8:54PM Jul 29, 2020
Hi, welcome back to hold 2020, we're really glad that you're with us right now we have the, the famous Russell Hansen was a well known bio scientist and bio hacker. So I'd like to pass it over to Russell and he's going to bring you his presentation straightaway. Thank you.
All right, Russell you're on.
Alright thanks JP for the introduction and welcome to everybody on the stream My name is Russell Hansen on my hair in the beautiful World Trade Center in New York City. Not in the hotel Pennsylvania though. I've always had some good times down there. So, my talk is about brain hacking and neuroscience and artificial intelligence, and let's get going.
So, just a little bit of background about me.
In case you don't know me already. The presenter said I'm infamous I'm not sure that's true. I'm not sure I'm famous either. So, Back in the day, vise wrote about me saying this guy wants to help you download your brain, I do. I'd love to help you download your brain. I've been talking about it for a long time on vice and the 21st chaos communication Congress in Berlin, and lots of other places. Pretty nice hacker conference is terrible called low card, and I worked as a professor at Mount Sinai for a while on the genetics and genomic sciences department where I worked on brain imaging and
applications of artificial intelligence to brain imaging.
I'm in the live version of this talk, I, I apologize for not being able to bring in some really fancy brain imaging machines, but on here we cannot view, you know, a typical MRI or CT or transcranial magnetic stimulation. We can also look at, you know, mice being hooked up to
different types of optogenetic
interfaces so you know you can
inject mouse brains with
reagents that cause the mouse brain to react to light, and also to allow you to tap into the mouse's brain. Using an implant in the mouse's brain, and here are the masses in a virtual environment so you can't have the mouse so working in a regular kind of maze environment you have to put them on this funny little ball. But this is a real experiment. This is done.
So, you know,
I believe that neuroscience and artificial intelligence are are really picking up and they have been for many years now, frankly, so one can use synthetic data from trained neural networks, you can use highly parallelized multimodal imaging on traditional radiology, including MRIs and CTS pet scanners Positron Emission Tomography not animal type of pet scanners. In addition, there's traditional medical imaging so histology and pathology, and the precision medicine scheme of getting the right drugs to the right person at the right time can be hugely augmented by using artificial intelligence so instead of a doctor looking at a patient and having to biopsy their brain, you know, a MRI machine with a trained neural network can simply scan that person's brain and determine pretty rapidly what's going wrong with them, just by using huge amounts of trained, or labeled neural network data. Sometimes people like this illustration of the similarities and differences between neural networks, and artificial neural networks and real neural networks so there's. See the soundtrack wasn't that interesting anyway. So, here what you see is a perceptron type neural network using the best handwriting database with 2000 hidden neurons and it's like 1.1 million signups. So this is a rather old fashioned type of neural network. And as this video progresses, you'll see it move through several different types of neural networks. From perceptrons to multi layer perceptrons, and so forth. And finally ending up spiking neural networks so humans and mice and rats and chimpanzees, all all all organic organisms use some type of spiking neural network.
Let's fast forward a little bit, so this last one is a spiking neural network, and then the spiking neural network you can see how things that you know the structure of the network is really random, and it's still able to perform the sun, and reading recognition task but you can see how this, you know, this is not, this is how a biological brain is organized, much like your brain or a mouse brain. Very different from the multi layer perceptron but it still accomplishes the same thing. So sometimes we joke in the lab that you know a scientist uses his or her brain to train a neural network to analyze brands which are themselves neural networks so you have a scientist, you have a multi layer input layer hidden layer and output layer, and you use that to analyze normal brains and Alzheimer's brain so you know there's no limits really as to how many types of brains can be used to analyze other brains. And we do this in the lab on things like Alzheimer's data so this is neuropathology data. These are little neurons, and when they're brown that means they're dying or close to die. It's a there, infected with a disease called tau pathology and the microtubule associated protein tau grows excessively when organisms are hit in the head boxers football players, and so on. And, you know, what we did is we built a little neural network, and what it was able to do is to annotate, this is a real human brain, a very old person. Most likely, pretty serious dementia, Alzheimer's. And, you know, we continue to do this, to this day. One of the ways you can do this is using a fully convolutional network to go from patches to pixels. So, an FCM or a fully convolutional neural network allows you to do pixel wise annotation of brains. And, you know, again this is just sort of the neural network architecture to do these types of calculations image to image translation is actually a pretty cool area of artificial intelligence right now so you can turn a horse into a zebra. And it's just sort of a trick. And what this means is that in a scientific or biological video, you can you can transform one dataset into another so you could transform a dimension data set into an Alzheimer's data set or a healthy brain to unhealthy brain, by using a consistent loss function.
And you can do it in real time. If your computers are fast enough. And you know, it's a little bit choppy. Maybe, but still pretty convincing sort of it, it looks like a computer generated zebra, like the lighting isn't bright, and it's it's a research project this is in Hollywood. Not many other different types of neural networks can be used for other types of transformations from different types of brain data, so you can turn a bright fields into a true neuron map, and so forth. This is all pretty scientific and I, I rather get to some other areas in the talk so I'm just skipping through these couple here. But one of the really cool things that I like is how you can use neural networks to do super resolution so super resolution means you can take, you know, an image or you can see one millimeter in resolution and bump it up to a micron. And in this case, a, you know, a three Tesla image Tesla's the strength of the magnet can be transported to seven Tesla. By using super resolution. And here, along the bottom, I don't know if you can see my mouse but you probably can, as you see how a three Tesla, can a three Tesla image cam using a neural network system be made to look almost equivalent to a seven Tesla ground truth image. Again, this is our outline data sets. So moving on to brain backups stuff. You probably noticed that my talk is called. What's my brain got to do with me. I got a man, that's an homage to positive K, and his album, the skills that pay the bills. Okay so, um so brain backups is a project that's been going for quite a few years to backup human brains and the goal is to make full connectome scale resolution brains and allow regular old people to back up their brains to computers and do what they like with them, it is your brain after all. And obviously you can do what you want to your brain, you can also do what you want your body. You know, back in the day before telepathy. People thought telepathy was science fiction on Celebrity those direct brain to brain communication is quite easy, actually. So in this experiment. One individual had a, an ag type headset and the other had a large magnet to their, their brain, and they communicated with a phosphine or sort of a bright light was displayed in your brain. Using a TMS coil. And, you know, in addition, the entire human brain has already been uploaded. You can check it out at Atlas stop brain dash map. org. This is a 30 year old woman who died. Somehow, and she took her brain to science and it's fully accessible on the internet right now. It's, um, I would argue that it does not have her memories or experiences, it's really images of her neurons, but it's conceivable that using some of the neural network technologies like I described in the last couple of slides that this information could be augmented to, to provide some of the data that's missing from these images. Right now, I believe they were, he needs teams
or not neuroanatomy images.
You know, just a brief overview of what is the brain backup. And what is it good for brain backups is actually a trademark name for the connectome. It's a network of neural connections in your brain and submitted metadata about those neurons. We see that the connector has many parallels to the human genome. So getting $1,000 genome versus the. million dollar connectome is outstanding. For us, in our work of brain backups and the scientific community as well. There are health applications, there are educational applications, there are technology applications there entertainment applications as you've sort of seen in many movies, starting in the 80s or in earlier. I would say that some of the more sketchy areas may be in the business area so if you if you know what types of brains are likely to to buy different products you can sort of do social media marketing in a way that is unheard of at present neuro marketing has been around for quite some time but the better. The brain models, the more money you're going to make. When I first started talking about this. There was a lot of skepticism you know it just sounds like science fiction I don't believe that supercomputers could model the human brain. And so back in 2014, almost six years ago now. The Telegraph, the British communication, wrote an article about, you know, what's the scale of acceleration that's needed to do real time calculation of human cognition, frankly, and that number is about 2000 times so computers were 2000 times faster. And it's, you know, of course, you can spend as much money as you want on like a soft desert or AWS or some other cloud provider, it's just a matter of money to determine you know how fast is fast. There are limits of course you know this is a. The biggest supercomputer in the world or one of the largest supercomputers in the world so you can actually scale that 2000 times
on the same computer hardware. Just as an estimate.
How's everybody doing out there I can't watch the matrix in my PowerPoint preview at the same time but I look forward to your questions feel free to send them along. I will do my best to answer them shortly. I think that it's interesting to to really show on a spreadsheet type format what a Human Connectome looks like. And the Human Connectome, or rather the Human Connectome, it has many things in common with C. elegans connectome. And so here is C. elegans connectome, this is not the Human Connectome. I don't have it in this compact form, but I wish that I did right now. Um, you know it's it's neuron one is connected to neuron two via a particular type of connection, and the neurotransmitter along on second connection is, you know, glutamate neurotransmitter. There are different strengths of these transmitters so there's two glutamate score glutamate seven glutamates are fine but it makes, in this particular C. elegans worm connectome. Some of them some, some of them received and some of them are gap junctions, and, you know, each of these origins and targets has a particular meaning and significance as I imagine a computer hacker conference, there are a lot of computer scientists and computer hackers and if you want to download a software that will allow you to
to use the 2015 version of the software.
There's probably it's probably updated since then. But that's the neural neural simulation tool is one of the best ones.
And this is what a
neural stimulation looks like.
So this is the C. elegans connectome
running through its motor output of the motor neurons connectome so there is a sonar neuron food neuron. And they update and provide stimulus for this synthetic organism to navigate this environment.
A lot of money has been going into
the space, you know, Brian Johnson along last knurling has done a lot of press recently. One question that's also frequently asked that I like to put into concrete terms is how big is a connectome and bytes. So, the human brain contains on average. I like the number 86 billion but it's about 100 billion nerve cells or neurons, on average, each neuron is connected to their neurons through 10,000 synapses. And you got about 909 terabytes. There's a lot of sort of hand waving going on there but in order to fit the calculation into a couple of lines, it's a rough order of magnitude calculation. So that means that, You know, based on a ballpark price of a terabyte of storage, you know, full storage for all human neurons, probably around these things is 10,000 bucks
or the price of a very cheap used car
to extrapolate out a little bit. If you have 500 inputs per neuron and the adjacency lists. You get two kilobytes per neuron, if you assume that there are about 1000 neurons subtypes, you get another 10 bits. And if you assume that each input sometimes has 1000 states you have another fun post, which would give you a reduced number, even though your model has increased in complexity, so to say. So, you know, we could say that it's between 384 terabytes and 909 terabytes. Also this information is highly compressible. So, again, you know, the numbers are going down with, with more information so 200 to 300 terabytes.
Russell. Whenever you're ready, we do have some questions for you from the audience.
Sure, I'll take your question right now.
All right. Question is, if too loud telepathic can be measured.
And it'd be recorded or played back.
Yes, absolutely. So, I mean,
in this particular toy example that I showed on the last slides, the direct brain to brain communication was really a binary zero and one. And obviously you can simply record a binary zero on one to a text file or a binary file, and replay them infinitely, so it Yes absolutely, you know, in, in almost any form or fashion, a telepathic information can be recorded to a computer and replayed instantly.
And with the headset that you have. I'm like a ham operator myself. And so it's interesting about the frequencies, because I'm sure the brain operates itself at one frequency and does the headset modulate the frequency or something to make it easier for the telepathic to go or is that all organic at the same brain frequency.
That's a good question. Yeah, I'm, I'm also, I have a yes to and about fan radio right behind me I am I like paragliding with my radios The suit was much better about things, 25 bucks. How can you argue with that. Yeah. It's a. There are many different frequencies, and wavelengths that the human brain works out there are alpha waves and beta waves, and there are types. Hi. So the it. It's very complicated but you know at some level there's chemical impulses that
to electrical potentials. As as neurons fire, and these in turn give us electromagnetic fields that can be picked up by, you know, a microphone or a speaker, and these are these are very faint signals. And so in an EEG headset is one of the most normal ways to read these signals and eg headset it's just a little cap with. I'll show on actually two slides. So it's really much. Yeah, yeah, thank you for the question. I appreciate it. Um, back to my slides. So you know every every organism has a different brain but it's organized in a different to do what that organism is designed to do, so to say, so you know, a rat has 200 million neurons, a mouse only has 71 million neurons and so we do a lot of experiments with mice with these very very small brains in the lab. So the old factories have is really large and a mouse or, you know, a large portion of the mouse's brain is just oriented in its snout for finding food and smelling its environment. If you as you go up the scale you know a copy of Berra has 1.6 billion neurons for a monkey has 3.2 billion neurons and down here is our brain with 86 billion neurons, and it's really fascinating to consider all of the different characteristics of these different brains I can't, I'm not an expert on the monkey brain or the brain or the brain but but each of these brands really does, you know, pharma set stuff or mouse stuff for rats. Rats are much much smarter than mice as you as you might infer from an having a three times bigger brain. Imagine if you had a friend whose brain was three times bigger than yours, they might actually be smarter. Anyway, onto a more brain computer interface stuff. So I was talking a little bit about these eg headsets or any eg headset, as a non surgical way of getting brain impulses from from inside the skull, to the outside and reading them on a, on a computer. And a lot of people, you know the resolution isn't very good because there are multiple layers of skin and bone and fluid between the brain and what you can get on the outside of the head. And, you know, a direct reading on the surface of the brain, on the, you know, the surface of the membrane on top of the brain is really much, much better, and the applications are much more rich, frankly, so in this example there's a 64 electrode by a 64, an eight by 864 electrode array with a built in antenna and titanium housing and hermetic feed through.
And so with this type of implant, you can really
one of the issues is the skull actually, it needs support to to maintain its curved shape, and so you have to you have to interface with the skull and provide like a, a support to the to maintain the sculpt the skull from really sort of caving in on the person or a monkey or what have you, so this this really has to integrate with the bone of the skull.
And you know, many, many projects and companies and research endeavors have been working on different types of eg headsets. This one, the amount of POC is one that was quite popular A while back, it's a wireless bluetooth headset with a rather poor resolution that uses sailing wedded electrodes, but it's still pretty useful and fun and you can do meditation games and kind of pretend that you can you can play Pong you can think to the left and to the right, train a neural network to read your brain, while you're thinking to the right or to the left. Open eg in Brooklyn is one of the more popular. Open projects or open BCI.
And we used to have.
So this is a gentleman here is using ag headset to control synthesizer.
Anyway, um so yeah you know sometimes people think it would be pretty wild if they could use their brains to control a musical instrument, and it's pretty straightforward. This is a case with the, the amplitudes of these different pads on this person's brain, and they are playing a musical instrument using the brain directly. Things we used to do for fun. Um, how do you make a Wi Fi connected brain. Well, you just need to implant, the device from the previous two slides and
right type, if config brain zero up.
Sounds amazingly fun. Many different projects have been envisioned by the federal government including DARPA to bring these things to life. And, you know, the. This stretch goal from DARPA was to record for more than a million neurons and stimulate more than 100,000 rounds in a single, I believe it was a cubic centimeter device. At the time, and still, there is no device whatsoever, that can record from a million neurons to nearly 100,000. There are a couple of ideas and sort of prototypes where they have a fiber optic bundle, and the fiber optic bundle can be used as part of the interface but frankly in order to get the fiber optic bundle working it needs to transfect the human neurons with some kind of floor for something and it's really just not a great technology making. But if you do have these direct brain computer interfaces, or brain machine interfaces as they're called. Having an exoskeleton or controlling motor cortex is seems like an obvious extension and you know, why would the Department of Defense not be interested in these things really.
using these as stents that is inserted through blood vessels in the brain is another relatively non invasive way to get electrodes into the brain.
This is called the Centrum, which is a.
I believe still in hyper phone, a hypothetical brain machine interface, and it grows into the, the blood vessels of the brain, and transmits information back down the wire. It's really a pretty fascinating building these things so the brain is essentially a entirely water. And so, radio signals work very poorly in water so if you have a submarine for instance you can't transmit radio from the submarine submarine has to trail a very long antenna So water is extremely attenuated to to radio signals and so a lot of the trick is actually to get around some of the basic physics of water in the brain, and how you how you could use other other methods to to sense things in the brain. Whereas connected mapping. You know, we believe that brain backups that you don't want to destroy the thing he wants to image frowny face frowning face. There are other hypotheses and there are other approaches to this one is knife edge scanning microscopy so in the knife edge scanning microscopy situation you, you have a diamond knife and
a microscope that are linked together.
And you can slice the tissue. It doesn't have to be brain tissue, it can be kidney tissue, it could be liver tissue but you know for our purposes, brain tissue is more interesting. But you know, frankly, this is a post mortem destructive. Highly invasive obviously procedure. Another interesting idea that's been proposed is using genome sequencing, to enable connectome sequencing. This is an idea that was proposed by Tony Zadar out in Cold Spring Harbor on Long Island, and the, the proposal here is that if you infect the brain with a virus or a synthetic virus that performs no harm. And each of these baryons has a barcode in the barcode will be transmitted throughout all the neurons in the brain. And essentially the positive there, if you could keep it from being degraded by nucleuses and proteases. Other other enzymes whose job it is to clean up junk like DNA barcodes, whose only goal is to enable complete connectome sequencing.
After a certain amount of time, you can simply sacrifice the organism, take their brain out and sort of grind it up and sequence, and all of the barcodes will indicate the exact structure of the connectome of that organism. again. You know it's clearly illegal to sacrifice a person for the purposes of scanning their brain but for mice and rats and chimps, you know, you just need to put a proposal together that's approved by your institutional IRB and you're off to the races. We believe that there are many alternatives, including artificial intelligence and nanoparticles that allow at least the opportunity to to perform novel non invasive in vivo imaging of the Human Connectome. Many of these imaging contrast agents can be bar coded. So if you have a bar coded nanoparticle then you can watch it as it diffuses around the brain, and one design is the login plus the contrast provided Asian, giving you a specific targeted contrast, or agent, and a other technology involves the use of target RNA aptamers with gold nanoparticles. In order to provide the same contrast. It turns out that this work that we were doing back in the day is still being used as we speak. For targeted coronavirus detection, so the antibody test actually uses something very similar to this the the antibody test is what's called a sandwich essay and a lateral flow device format where antibodies. Use these, these same nanoparticles with magnets attached to them to to bind to a test line on a lateral flow device. So, no, I would say that home diagnostics arc. This is like a home diagnostic, whereas today you can a woman can determine whether she's pregnant or not by using a diagnostic This is essentially a diagnostic for different forms of neurodegenerative disease.
Numerous studies have been done to decode what people are seeing in the brain, using artificial intelligence or simple line, simple logistic regression or something like that so using an MRI signal, one can get an approximation of what the person is looking at these neural networks are trained very simply by a person sitting in a MRI machine, and being presented with a clip. And then they take away the training data, and they reconstruct when a person is looking at another environment. I'd love to see these experiments updated. With modern techniques. This was the original experiment decoding the visual cortex, as well the the auditory cortex has been decoded all doesn't work well at all. Structure
town. Oh. Oh. Doubt
You know, this uses a one of these ecog arrays so that these little red dots are electrodes that are touching the surface of the brain. And so, doing this type of auditory reconstruction is not possible without really direct contact with the brain so you know no one is, is reading your thoughts without you really knowing that they're in there, touching your brain, so don't worry. But yeah, for scientists, this is,
as I mentioned, you know,
one powerful approach is the use of contrast agents here are these contrast agents, these little black dots are gold nanoparticles with glycine receptor ligands attached to them. The goal really is to augment, some of the earlier images that I showed from the brain Atlas, with more information about different subtypes of receptors and transporters. And these, these can simply be listed on, so you know, there are many different types of dopaminergic receptors. So, do you want D two D three d 45 so these are five different types of dopaminergic receptors, there are many different types of aggregator receptors, there are many different types of gabaergic receptors but they all actually have a gene expression ID or unique product session ID. And what that means is that they can have logins designed for those specific receptors specifically. There's no, there aren't very many unknowns. In terms of, you know, the structure of individual receptors it's really a matter of getting that information out of the brain into some format in a highly parallelized fashion.
A typical MRI or CT machine images that about two to three millimeters resolution. And, you know, getting to a micron or something like round really owes an outstanding goal here is a high resolution CT image of a piece of human brain with a very short scan time. So this was only a six minute scan but it was able to image 10,000 times better than traditional MRI and CT and zoom zooming up more here's even higher resolution neurons, and roads and some more gratuitous brain imagery and these are some data that we took on a high resolution nano CT machine, a bruker, and Octa machine. Here the voxel size was 20 microns, but that's not because we needed to. I mean, the machine image was down 2.2 microns, but in order to get this is a, this is a rat brain with a contrast agent, so all of the whiteness rules use blood vessels and a rat brain with a contrast agent that consists of gadolinium. And, you know, it's, it's pretty amazing to realize that right now there are machines that will image, a block of human tissue human tissue or actual shoe or whatever issue at the resolution we're seeking which is, you know, a 10th of the micron. It just requires a little bit of know how in terms of contrast to get in just to show up properly. So this is a healthy rat brain, and this is a diseased rat brain so these rats were part of an experiment we did where they were exposed to blasts, and so the blast exposed, like a person was in a tank and the bomb went off. You know that's that's the scenario that we're looking at here so if you, God forbid know somebody who's been a bomb attack. And you know there was an explosion very close to their head. This may be what their thermal vasculature looks like so very a huge loss of one foot so again this is the healthy rat brain. This is the blast exposed rat brain just you. Incredible pathology. Many, many people working in this area are quite enamored of the idea of nano robots. And so here is a an artist's depiction of what a nano robot might do were to cruise around with the brain and try to determine what is the possibility from neuron one so they're on to or they're on 50 million to neuron 60,000,001. This is not a particularly scientific depiction here, but what it shows is something similar to what I showed on the previous slide, which is a, a gold nanoparticle with dimensions something on the order of 25 to 45 nanometers with a leg and attached to it chemically. And that, again, has specificity to different targets and integrate.
Again, you know, artificial intelligence is a huge advantage here and so this is a block. Nine microns by six, seven microns by 4.9 microns, of the human somatic sensory cortex, and the goal here is to pull out all of the sinuses, so the sinuses are these flat sort of pancake like structure structures in the brain, and a neural network or a modified neural network was used to pull these off. So, these are the several hundred signups is a very small nine micron. Walk of human SMAP sensory cortex tissue. You know, there are a lot of people who, who think that, you know, there's something inherent about the brain that makes it impossible to simulate, or, you know, that it's just too complicated. And, you know, there may, there are many arguments against that and I think frankly over the many years I've been doing this, it's becoming more and more easy to disprove this but briefly, you know, brain functions are not computable computability actually has a definition of our mathematical definition that definition, essentially says that if there's a finite amount of information but it can't be computed quite easily. So, obviously, the human brain has a finite amount of information it's encompassed within your skull and there's a border, so it's it's finite. What are the ethical implications of brain imaging, you know, frankly, as a scientist and someone who works in hospitals with trying to improve patient care. You know, it's like if we knew what was wrong with with people with dementia or Alzheimer's or other mental health issues, then we'd be able to treat them more accurately so you know right now. I think that the human brain is one of the most important areas for, for science, and it's. The more information you have about the brain, the better we'll be able to treat people.
Now, in the future, you know,
brain imaging can save us from the odds, or at least it's worth trying. So I coined this term the Russell point. So in the Russell point human characteristics are preserved by necessity for the human performance first as computers get better than people. Hopefully the Russell point will be the point at which human performance exceeds machine performance and keeps us human. What's next, of course, there's a lot of work from companies and DARPA, and planning CPUs databases of neural codes. Elon Musk kernel and many other companies are working on that. I think that using artificial intelligence to analyze the brain will be hugely impactful in the next several years. In, frankly, optimistically making $1,000 connectome seems
somewhat easy now compared to when we started this.
Anyway, that's my talk, they telling me to get off the virtual stage here in New York City. I'm open for questions.
All right, thank you very much Russell that was really excellent and I My apologies for Miss speaking before I should have said well known and forget all that other stuff that I spoke about. All right. Our first question for you is, at hope we've had some talks addressing machine learning and AI as you were just mentioning about. They are so different today than what do you, then what you're talking about a sense of connectome. Can you comment on the utility of MMA modern AI techniques, relative to what you're working on.
Yeah, yeah, absolutely that's that's a big area right now. You know there are conferences on how to use brain imaging to augment AI. It's, it's, it's kind of a iterative loop right now so when we were doing these different studies on using AI to analyze brands, the biggest bottleneck was having a human annotate the brain because you know only a human knew what these different features were. And so I think having self training neural networks or you know whatever the next advances in deep neural networks will will make it easier to image the brand and then there will, there may be some kind of loop regarding, you know, a neural network that trains itself to analyze human brains in ways that humans couldn't do. I hope that answers your question.
Yes it does. Thank you very much, um, the next question that we have is, what about the issues of full simulation of these scans, like would it be ethical to try and wake up a brain in some software event. I think it might be not according to the questioner.
Yeah, yeah, not a question that comes up quite often, you know the paradox of waking up your, your brain back up and having them meet you. You know I think over the years the opinion on this has changed somewhat, I feel like it's sort of the great unknown, it's like, you know, before we had radios, wouldn't it be really was like talking to you and your radio but they were five miles away I mean that's that's just not possible. So I think, I think it's more of a societal or a climatization issue than it is really a technological issue I think the technological issues, relatively straightforward.
Yes, the evolution of things happens more and more rapidly it seems. So the next question is, are we closer to having a functional computer computational model of the brain or being able to build synthetic biological models of the brain.
Also, relatively common question, which is fine. It's a great it's a great question. Personally I think that building. You know, wet brains with real neurons at the level that that is interesting for any computation or, you know, consciousness is really very complicated and slow. So I think that you know having to grow in a layer by layer with neurons and having a face, and then putting them into a body so that it can experience life is not a particularly appealing technology to me I've just seen. If you've ever grown cells in the laboratory, you know, the idea is a little bit absurd. It's just building at 6 million neurons and then getting them into an environment where they were just isn't going to happen so I think that it's much more realistic to image the wet arounds and transfer them to accomplish a computational environment, and
and work with them. That's just my opinion.
I'm not, I'm not hearing your question, it seems like your videos, bro. Yeah, looks like you dropped out momentarily I'm I'm standing by let me go ahead and redo the next one. This is what about issues of full simulation of these scans to be ethical to try to wake up a brain in some software that and I realized you just said that wasn't that interesting to you but I think this is a question about the ethical implications.
I think ethics are a personal thing. So, you know, to me. Yes, I understand that there are questions as to whether this could be done but I think that there are yeah it's it's. Yes, it's important to understand what should be done versus you know whether it's technologically feasible i guess i don't i don't see the reason to do it. And so for that reason I I have trouble thinking of you know the ethics of doing it. Yeah, it's, it's hugely questionable I guess.
This last one is we're just in our last few seconds here and welcome back JP, just in the last few seconds. Will we ever be able to have a full non destructive way of scanning our own brains within our lifetimes.
I, you know, I, I'm obviously biased here I've been working on this for quite some time and I think that the
barriers are coming down at
to recognize features and you know automatic resolution from a millimeter to a hundredth of a millimeter by just using a neural network is is you know, it might be the missing piece puzzle of how to, how to image. A human brain on the 10th of a micron resolution which is approximately. The smallest feature you may need.
Well thank you very much, Russell Hansen backing up your brain. Here at hope 2020 Thank you very much. Doctor really appreciate you sharing with us and on behalf of all the volunteers and all the attendees, really appreciate you being here.
My pleasure. I've been trying to hope for. I started going 1012 years ago and it's always a good time. So, thank you for volunteering. Like.