Focused Ultrasound & Longevity using AI Professor Vincent Ferrera & Siraj Raval https://desci.nyc/transcripts/focused-ultrasound-longevity-using-ai Automatically generated transcript. May contain errors. [0:04] okay cool well um thank you guys for all coming to uh D NYC here and uh it's a pleasure to have you here and um I'll just tell you a little bit about like what is D NYC um it's a monthly Meetup that we talk about all different things about the intersection of blockchain and Science and uh we get a lot of people we have a site um it's about speakers it's about Community it's about pizza we record all our talks and make them accessible online um we have a group chat application here where you can like have like a little like after the um party is over you can still continue to [0:46] uh chat with people um so you can fill that form out cool um here's some of our previous speakers we started off with like a distributed mushroom collecting and then this guy who was doing self-administered psychedelic experiments and then someone who talked about bees and then we had Martin scy with AI in medicine we had the guy that invented the nfts we had the person with from research Hub who started that um we had this side combinator demo day we had um people on longevity from codow from age one we had a professor from um Albert Einstein medicine on longevity and tissue [1:27] replacement and um if you guys took pict pictures um make sure to tag us on Twitter or Instagram it's really we love it when people uh retweet it sort of sucks to uh have to beg for it but whatever um it was very helpful if you follow us and uh if you took pictures we I'll I'll retweet them and would greatly appreciate it can you send this point out afterwards so we have are the talks recorded yes should be here they recorded yes it is okay where do you uh put them um on D DCI do [2:10] NC okay um so take pictures tag us um so our what is decentralized science um it's like the emergence so it decentralized science came from uh sort of Defi and defi was like everyone was trading tokens everyone's creating tokens and exchanging value but what's really cool about dii is that people are trying to use the blockchain instead of just sort of swapping tokens people are adding data on chain people are doing analysis collaboratively people are splitting up the IP between people that started projects and so there's like exogenic information coming into the blockchain and then we're using like this distributed Ledger technology to [2:51] try and um make it more of a meritocracy around who's credited For What and incentivizing scientific research in new ways and there's all sorts of reasons why you'd want to do this like um you know traditional science has a lot of problems in in funding [Music] um in funding like there's like there's only so much funding to go around and it's typically granted by a government agency and uh decentralized science offers a way to crowdsource the funding uh peer review can sometimes be slow and distributed Ledger Technologies can offer a way for people to um crowdsource um um uh pure [3:32] Publications and um I think one of the more interesting things is like um allowing people to incentivize science in new ways So currently science is incentivized using sort of uh Academia and you have to get a postto or you have to get a professorship or you're trying to get into a good graduate school but and then you you didn't start a company but all these sort of incentives can sometimes be a little bit misaligned and create uh Mis and create sort of amalgamations that you weren't expecting to begin with what's cool around decentralized science is like uh you you're you when you start a project you're you're creating like sort of like a company and that company holds IP and [4:13] then everyone that sort of participates in the company um it gets tokens in reward for the data that they contribute in rewards for the research in the rewards for like the um uh numerical analysis that they do do work writing it up and those tokens Can then be used to uh buy whatever or as a status or for lots of different things so it makes the like the the beginning to the end process of doing science a little bit more open and a little bit more clear from the GetGo um so I'll just talk about and then for our next talk so we're going to do [4:55] another one and this one came out relatively quickly so this one's on uh February 22nd and we have the founder or the president of uh the enhanced games so the enhanced games are a really interesting idea of using um like the Olympic type Sports in competition and it's like the Olympics but instead of the Olympics being where they check people for doping very carefully it's like un un there's no checking you're supposed to dope and uh and uh what I think is interesting about it from a scientific perspective is um it's like a way into looking into performance enhancing drugs can help [5:37] with medicine or help with sports and right now medicine is very a little bit retrospective looking where it's like someone gets sick and then they come into a doctor and they're like okay we have all these different things but what's interesting about this is they taking a healthy person and either making them even more sick and get cancer at 25 or um trying to enhance their life uh and maybe maybe prevent death give them more muscle mass and sort of seeing what happens so I think it'll be interesting to see what happens uh where where they go with this and to look at it from a point of from the point of view of like a science perspective so uh the link to do that is here um does anyone have any questions [6:19] around decentralized science yes I guess uh would you mind comparing and contrasting it to AR yeah so archive is around the final publishing of it and there you're still not you're not getting any like you've done all the research it's been funded by the government it's been funded by uh it's the research happens within a university through a pi through graduate students through research assistance um they then send out surveys to people that don't necessarily uh care about taking the survey and are just trying to get through it as quickly as possible to get their $20 Amazon gift card um and then they uh go take all the research up and [7:01] write it and publish it an archive where there's either no peer review or uh Open Access quickly but you don't you can't see like necessarily who contributed what to the science project and uh if you contribute you don't get any sort of credit Beyond like your name at the top of the paper whereas but it is a good step in the right direction whereas maybe a decentralized science approach would be someone comes up with a hypothesis that people can participate in and then multiple people add their data to the blockchain so there's Prov for every part of the data um everyone that contributes as part owner of the IP and Equity of the project and then they can sell their IP or Equity um for money for use it as a way to get a job or any [7:43] other number of things anyone else one more yes like my question with decentralized SCI that I always wonder is like when citations are the primary driver of funding currently what do you think would incentivize like a promising young scientist to say I'd want to contribute to a decentralized project if it potentially like takes away from the method of growth that are career currently yeah I mean I think you can make more money in decentralized science because you might be limited to a certain salary and so there might be a possibility of making more money in a uh unbounded way where you're sort of [8:23] working more for yourself as opposed to working within a traditional academic setting of course it comes with you can make no money as well um and it's for people I think it's not trying to compete for those types of things honestly I think it's for people who are looking to go a separate path in trying to do research as opposed to um trying to compete with things so like decentralized science projects aren't the type of thing that um aren't people aren't trying to do like basic research right now whereas in like traditional universities there's a lot of basic research being done but there's also a lot of non-basic research being done in universities which I think can be better done in the decentralized science approach than through the traditional [9:05] Academia cool okay so we have um we have two amazing speakers tonight um the first is Vincent who uh is a professor at Columbia University in the Neuroscience department and uh he's been doing this very interesting thing which we're going to talk about on focused ultrasound and focused ultrasound has been uh as of late become very very popular because it is completely non-invasive and it's uh pretty generalizable around the body and very cheap to administer so uh it'll he's been but he has been doing this for 12 years so we're excited to uh hear what [9:46] he has to say thank you and we thank you for coming out tonight as [Applause] well I think it might be better yeah let me just make sure it works to get to Stage okay all right okay um well thank you very much for inviting me and thank you all for coming out this is really exciting I fully support decentralized science we um you know we're putting a lot of data sets on the internet and you know publishing on bio archive and stuff like that um it's a really great concept um and you know [10:27] freeing ourselves from the shackles of traditional science um you know can only benefit everybody I think um also I have to say this is the first time I've ever given a talk in socks and that's perfectly fitting because this is going to be a little bit casual so um you know bear with me a little bit uh so yeah that's me um I'm at Columbia uh my disclosures are I'm just a poor scientist I have no Financial relationships um who is this guy uh so I'm a cognitive neuroscientist I'm interested in learning memory attention [11:08] decision- making in prefrontal cortex and in the basal ganglia um of primates uh which includes monkeys and humans um the tools that I use I heavily rely on behavioral methods electrophysiological methods uh these neat new neurotransmitter reporters which is a whole other story that I'd love to tell you about but no time uh chemogenetics which you might know something about and what I'm going to focus on uh no pun intended is focused ultrasound and so you know at this stage in my life you know I'm a full Professor my kids are grown I'm an empty nester I have a real personal interest in this topic of healthy cognitive aging um throughout [11:50] the lifespan so from childhood to you know Cradle to the Grave as they say um and the field as a whole I think um you know we we love our tools we love our optogenetics and chemogenetics and all this invasive kind of stuff and deep brain stimulation but you know we really as a colleague of mine said nobody wants a whole drill in their head right and we want non-invasive Technologies and ultimately you know a 100 years from now or you know star date 2347 whenever it happens um we do want to get to a fully [12:32] non-invasive kind of you know Star Trek sick Bas scenario so that's you know I hope you know we're taking some small steps in that direction now so that's a little bit what I'm going to talk about so we're talking about ultrasound what is ultrasound well it's really any sound frequency above 20 khz 20 kerz is the limit of human uh auditory perception and so we just sort of arbitrarily say that anything above that um but the kind of ultrasound we're doing is actually uh 10 times higher it's in the 200 khz to 2 megahertz uh range generally um and the way we produce that is we [13:14] have a Paso Electric device that's embedded in this ceramic it's basically just a shallow ceramic Bowl um and this can um do two things one is it can produce very high frequencies like I talked about but also it can focus the ultrasound down to um we like to talk about the grain of sand or sorry the grain of rice whatever so you know we talk about the focus being about the size of grain of rice maybe it's a little bit bigger you know um that's something we can talk about um so this is you know different but complimentary to the traditional Imaging [13:55] B mode ultrasound that you're all familiar with um um and so what does um you know what does ultrasound do well lower frequencies you know so ultrasound can penetrate into the brain and the reason it can do that is because the pressures at the face of what we call the transducer so that little bll shaped thing I showed you we call it a transducer the pressures at the face are very low and then the pressure increases as you get towards the focus um we can vary the frequency low frequency ques have deeper penetration but a bit uh you know the focus is more spread out higher frequencies have less penetration but a [14:36] smaller uh tighter Focus so T we tend to use higher frequencies in smaller animals like mice and lower frequencies in you know primates and humans um what are the effects on biological tissue well basically I mean there's a mechanical effect of course it sound is a pressure wave and and it's um you know causes vibration as we all know there's there's also a thermal effect so there's heating at the focus because of the energy that's being deposited um and something that you probably might never never have heard of is something called um the radiation Force so the sound wave actually pushes and causes a displacement of tissue and that's [15:18] something that you can actually image you can image it with MRI you can also image it with ultrasound itself uh like I said on the B mode Imaging and so the the displacement can give you sort of a readout um of the pressure wave inside the vessel in this case the skull that we're introducing it and so that's really exciting because um I mean we do have this thick skull to deal with it's it's a bit of a problem but we do have the sort of crazy idea that we can have a completely um self-contained system where we have focused ultrasound um going in and and [15:58] stimulating and we can read out with imaging ultrasound um this is sort of Cutting Edge stuff um but this uh this field goes way back so the earliest study that people generally refer to was uh by Harvey in 1929 almost 100 years ago and he just did the you know the Luigi galvani experiment basically of causing a frog leg to Twitch with ultrasound and then he showed that you could block that with kurari so it was mediated by the Nerds um then about 30 years later um Fry at all did an experiment um in a [16:41] living animal where they uh focused ultrasound into the uh brain stem in particular the visual Thalamus and they were able to measure electrical potentials on the surface of the brain the visual cortex okay so that's some evidence that the ultrasound was actually stimulating the thalamus and stimulating the you know the fibers that go from the thalamus to the cortex and causing a response um and now we can actually uh induce what are called phosphines the phosphine is just a flash of light that you see um you know sometimes when you're laded or when you get hit on the head you know that can be induced um with [17:23] ultrasound uh so I'm going to talk about basically you know if you're at the field there there are basically four ways that ultrasound is being used um neurom modulation which is a direct mechanical or thermal effect that directly stimulates the brain uh bloodb brain barrier opening which I'm going to talk about where we combine Ultrasound with intravenous microbubbles um then there's a thermal oblation which is a purely thermal effect and is actually FDA approved as a treatment for essential tremor but I'm not really going to talk about that a lot of uh medical centers now uh have these devices and are doing ultrasound [18:04] ablations as a service people come in with a central Tremor they have the procedure they walk out without a central Tremor it's really amazing but I'm not really involved in that research U and finally functional ultrasound is a way it's a novel Imaging modality where you can actually image blood flow so you can actually image like an MRI you can image the function of the brain in real time and people are using this really cool ways like to read out um you know movement and tension and things like that and the parial cortex um but I'm going to talk about mainly these first two um in terms of the [18:45] biophysical mechanism people are interested in uh these uh trip channels um which are present in the brain and uh are temperature sensitive and probably have something to do with the response we don't really know lot about this or I should say I don't know a lot about this it's not my area of research but this is something to um keep an eye on in the future uh this is what you know got arum PTI and the Nobel Prize a couple years ago um okay so I'm going to talk mainly about you know not biophysical but the whole other end of the spectrum we going talk about behavior um in primates and how that can be modulated with ultrasound and specifically you [19:25] know how we can you know you know change their uh cognitive performance on certain tasks so I'm going to show you uh so this is the setup we have a monkey and he's uh interacting with a touch panel and he's doing a little task he touches the panel and um you know he gets uh he gets apple juice when he does it right he doesn't get apple juice when he does it wrong that's very motivating and he loves to do that and he'll sit there for a couple hours every day and so our first question was okay if if you have an animal in awake animal what is their reaction to Ultrasound when you apply it to the brain um you know we [20:05] didn't know anything could happen right I I can't even go through all the possibilities so the first thing we wanted to see is is uh you know the first question is what happens really just pretty basic so we've got him doing this task and I'll show you exactly what he's doing in a second here okay so here's a little move and um so there's these two motion patches and one one of the patches has motion that's a little more coherent uh than the other and you can sort of sense what the direction is so uh this one is moving uh coherently to the right um and [20:48] the other one's just kind of random so he has to pick the patch that looks like it has more coherent motion and they get really good at this and you can make the task easier and harder just by varying this parameter okay um next slide uh we're targeting um the basil ganglia specifically so this is uh a slice like this it's an MRI slice what we call a coronal slice um this is all cortex um it's not a great MRI so I apologize for the low contrast U this is what we call the striatum it's a part of the basal ganglia um that we think is involved in learning attention [21:28] memory and decision making um and divided into two structures the kodate nucleus and the paman um and that's as much neuro Anatomy as you're going to get so um and then this is a simulation of the pressure wave inside the brain that we're hoping to produce so the transducer is sitting here and this is um this is the pressure and so here we're targeting the structure called the CATE nucleus and here we're targeting the putam and so um so the subject is sitting and he's doing this task and we give him um a few minutes to warm up so this is kind of like the warmup period and he just gets [22:09] into a rhythm of you know hitting the touchcreen and getting his juice and then this is where we turn on the ultrasound and we turn it on Just for 2 minutes and um this is at 500 khz and it's pulsed so we don't want to you know we don't want to cook his brain and we're using a low pressure um so the pressures are on the order of 200 kilopascals to 1.2 so what's a pascal uh one pascal is uh 1 Newton per meter squar um 100,000 or yeah 100,000 pascals is roughly one atmosphere for people who are into physics okay so that's that's as much uh it's it would translate into [22:50] 194 debels if you could hear it but you can't because it's way too high uh but that's like a rocket launch basically uh so it's a pretty high pressure um and so you know we're worried about what it's going to do to their brain obviously but um you know the animal started working we turned on the ultrasound for 2 minutes um it didn't bother him he just sat there you know punching the screen and doing his thing you know as if nothing was happening and we checked you know to make sure that we actually turned the thing on you know and and did all the kind of Sanity checks do and we measur the output and this and that and indeed the [23:31] device was on but the uh animal kept working um and not only did he keep working but he actually got better and better um so this is just showing his response time um and his response times are actually getting faster after we turned down the ultrasound so that was actually a good sign it was like the first thing we ever did uh with this and we were like I said we were really worried we didn't know it was going to happen if be there'd be some weird reaction or or whatever so so this is a good sign and you know we felt like we're kind kind of on the right track this task gives us a lot of things we can uh measure so we can look at motivation um in terms of response times and things like that and we found that [24:12] the ultrasound actually improved its motivation when we turned it on we found that his performance was better in terms of he had more correct responses and he was faster at producing those responses so the you know we were enhancing his his uh his processing speed a little bit by turning on the ultrasound so that was interesting um that's promis we did this experiment so clinicians always you know get on my case because basic scientists like to do you know they like to do proof of principle and just get stuff you know published as quick as possible and [24:52] long-term studies just aren't done and clinicians really hate that because you know it turns so you know stuff will get approved for human use and it turns out the long-term studies were never done and and it's a problem so we did a long over a period of two years um we did 140 um sonications in two animals and then we looked at their MRIs and their MRIs were um as if they had never been treated um basically there's no difference in the pre and the post MRI after two years and 140 of these treatments so again that's that's a good sign it means we're you [25:33] know we're doing we're having an effect on Behavior we're not destroying the brain and you know that that's kind of important if you're going to do this stuff in humans uh we're also interested in memory so here's a little memory uh experiment and see if you can figure out what's going on here um I'll give you a minute [26:09] okay so anybody get it yeah what what shape St throughout each trial right right so there's an array of shapes and then there's a little delay and then another array and one of them is changed and you just have to figure out which one change um and it turns out that's a fairly effective way of measuring what's called working memory or short-term memory um and uh so we did that and we had a control condition where there was no memory requirement and um these are the controls the Shams without ultrasound and this is the error rate as a function of how long of a delay there was between the two things and we you know we [26:49] measured the Baseline error rate and then with uh with ultrasound and we found that the error rates went down actually in both conditions um so we think that uh you know possibly you know there could be some benefit uh to short-term memory of doing this um do you know the underlying cause that's like is it increasing the blood flow in the area or like so we think we're just it's basically just stimulation so we we're we're actually doing experiments now to compare Ultrasound with you know basic like deep brain stimulation got it um and so we think we think but we don't know we haven't really measured the neural neural activity which is [27:30] something we have to do but there when we get to the end I'll show you another possibility which is um that we're actually altering functional conductivity in the brain okay um which we've shown with MRI so the next thing I'm going to talk about so that's just ultrasound alone next is we're going to combine Ultrasound with microbubbles and use that to open the bloodb brain barrier and the way it works is that we have these microbubbles they're little um they're little uh lipid spheres they're filled with a gas and they're about five microns in diameter five microns turns out to be about the diameter of a capillary and so when they go through the capillary um and they interact with the ultrasound they just [28:13] oscillate a little bit and that oscillation puts a little bit of pressure on the capillary walls and makes the capillaries a little bit leaky because the capillaries on the brain have very tight uh uh Junctions in their endoth cells that prevent uh things like you know things that are dangerous like poisons and toxins and and viruses and so forth from getting into your brain um but we can use this method to um to to make the capillaries a little bit leakier this was first done in mice and then my lab I think was the first to do this in primate model um but now it's done in humans and so forth in their clinical trials and so forth um this is [28:55] actually showing um this resolution isn't great here but this is a little tube filled with micro Bubbles and this is showing the ultrasound pressure and you can see when the pressure comes on right there you can see that change did you see that okay so we'll go one more cycle okay ultrasound comes on there and you can see the little bit of expansion of this is just a plastic tube it's not a real capillary but um that's the basic principle so again you know the question is is it is it actually safe to do this or you know does it even work at all in a large [29:36] animal I mean it works in mice but mice have a very thin skull uh humans have a fairly thick skull and not only do we have thick skulls but are you know every human has a different shape and thickness to their skull and so there are a lot of challenges there um so again this is the pressure wave simulated and then we combine this with the micro Bubbles and and what we found was that uh we did a contrast enhanced MRI and we found that the blood brain barrier was um you know permeabilized a little bit in the regen of the highest uh ultrasound intensity um and then that's here and we've done this now uh I [30:17] want to say hundreds of times um we've done it a lot uh and it's extremely reliable and you know one of the main things that we're working on is is just making it more precise and and um and uh controllable anyway this is just showing this is this is a penny these are some openings um and you can see the size so we can control we have pretty good control of over the size you know by varying the pressure we can you know choose different targets and so forth and we can uh construct you know actual physical models of the pressure wave inside the head and predict what's going to happen you trying and um Target like [30:58] density as opposed to volumes so if you have like a resonant frequency you could possibly oscillate some density yeah it's an interesting idea so it tends to um it tends to work better in Gray matter than white matter uh we don't know exactly why it could just have to do with the amount of capillaries or it could just have to do with the structure of the tissue but I mean it's a good point we we need to have simulations that take into account variations in the structure in the actual tissue structure and you know whether it's fat or neurons or whatever um could like Target a certain like type of thing right like if you hit the [31:39] resonant frequen if some if there was like a really like a piece of paper is going to have the same resonant frequency as like a as a yeah it's an interesting idea I have to think about that more um how we could do that it's I don't know if the tool is really that precise um anyway people ask how long does it stay open so we did an experiment and we measured you know we opened on Day Zero and then we measured um one days one two and four after that and it took about two days for so so when you open the blood brain barrier actually you're you know you're causing the vessels to become leaky um you're also causing a little bit of inflammation U but that seems to to [32:20] close up um within about 48 hours which is good and bad I mean it's it gives you window sort of a therapeutic window uh to work with um because you're going to have the barrier that's you know it's open for a period of time where you could you know do the procedure take the patient back you know to their hospital room and then give them uh drugs and and really monitor monitor closely what's happening um but it doesn't stay open uh indefinitely so um that means that you know if you're going to do this therapeutically you'd probably have to do it repeatedly and then that raises some real safety issues that have to be [33:01] dealt with um again we did this um and we didn't know what was going to happen uh you know in an awake subject um but we did it and we got the same result that the uh in this case um we could apply the ultrasound and we could give the micro bubbles intervenous uh open the bloodb brain barrier and it didn't disrupt the animals Behavior at all in fact the animals performance was a little bit better on days when we disrupted the blood brain bear here and we don't know we don't know why but um they don't they don't react to it it's not painful it's [33:41] not um you know it doesn't cause disorientation or anything like that which is all good things and and stuff that we really needed to know before we started doing this in humans um and then uh so in terms of mechanism I mean we're really just scratching the surface right now we don't know what's going on I mean when you open the blood brain barrier you could be changing um oxygenation of the tissue access to glucose access to electrolytes uh we're we're looking at these things using some MRI methods so we're looking at oxygenation using something called cuold which is a quantitative uh form of MRI [34:21] to look at oxygen levels in the brain um but this is another Finding um that that came up which is that uh when you target one region you can actually change the functional connectivity of that of the targeted region with with other regions so the ultrasound can actually you know so so functional conductivity is it's a you know it's a method to look at correlated activity using MRI um and you can sort of you know tease out networks and so forth and so the ultrasound can actually change the functional connectivity I want to be clear we're not changing the actual structural connectivity of the brain we're not [35:02] growing axons and things like that but we are changing how brain you know remote brain regions interact with the region where the focus ultrasound was targeted how long does it take for the change to uh we haven't actually measured the time course but typically we'll do the ultrasound procedure wait about an hour and then do the MRI and we can see the changes and if you don't uh stimulate again do the changes stay uh good question we haven't done that experiment exactly um why wouldn't you expect some heavy in idea of like the axons Android firing together firing together more more intensive than they [35:44] did before uh that's that's a certainly a potential mechanism for this um yeah I mean that's a good idea we would you know then want to go and actually measure that you know physiologically and and see if that's the case are the results path dependent like presumably there's some gradient of action potential that's happen as the wave coales and become stronger so if you target the same area from like different angles like you still see the same results oh that's a good question we haven't actually tried that but uh it affects different networks differently um we know that you know we we published papers looking at you know the default mode Network the you know attention Network and so forth so um [36:25] they're affected differently this is you know this is early days and um you know actually figuring out what networks are affected and how they're affected and and all that kind of stuff um you know there's a lot of work to be done um you know a long road ahead of us um but it's you know it's exciting this is actually a connectivity um map I forgot what they call it but um uh you know and we we've done this we've shown changes in functional connectivity both with ultrasound alone and ultrasound combined with the microbubbles um another thing that we found uh was so this was all this is all [37:07] done without drugs and and one thing that we're interested in is using uh this technique for targeted drug delivery so we did some experiments um I'm not going to show the videos but we did um uh we targeted the cod8 nucleus we opened the blood brain barrier uh we administered a drug called down paradon which is a dopamine antagonist that's used for nausea it doesn't cross the blood brain barrier naturally um and so when we open the blood brain barrier we saw an attentional deficit um which you know we weren't that happy about getting a deficit but it it showed that the drug was working so this antagonist could [37:48] actually um cause uh an attentional um uh you know it's called hemilateral neglect and it's basically it's a very common uh commonly seen after right prial Strokes you'll have patients that sort of just ignore the left side of space they don't Orient to it visually um their body representation has changed and so forth um and we saw a a very clear transient example that when we use down paranome we also used a serotonin antagonist called andanson again this is a nausea medication to taken by C cancer patients uh when we gave it to the [38:29] monkeys after the blood brain barrier opening they had a little bit of a neglect syndrome um and it went away after like 12 hours or something what is uh so it's a it's an attention syndrome where um the subject basically doesn't Orient to one/ half of space um so they will if you you know if you show them a piece of food in the um intact Hemi field they'll reach out and grab it if you show it in the impaired Hemy field they have no reaction um and this happened during and after the BBD [39:10] opening yeah okay so there was improved performance in some ways and then dramatically worse than others right but using the these are antagonists so presumably if we used Agonist we might get the opposite yeah beside our um enhancing delivery of the drugs in Alzheimer disease did you do any testing on delivery on drugs and also the antibodies do you any test on the antibodies as well so my colleague has a clinical trial on Alzheimer's uh happening now and I think they're looking at some of those things um uh you know we don't really have a good [39:50] primate model of of Alzheimer's right now so uh you know that that were is mostly being done in humans okay um and I think um so I'm happy to talk about future directions I don't know if I it okay um so you know the future of ultrasound in terms of um research I mean there's just a huge number of avenues that are being pursued um my lab is interested in optimizing the efficiency and the safety of the method um we're doing for example targeting uh based on individual Anatomy [40:33] because um you know human heads come in all shapes and sizes and skull thicknesses um and we're actually using um AI methods to try to predict how that's you know the skull is actually going to affect the pressure we inside the skull um something we're not really we're sort of thinking about not really doing but this is really cool it's called acoustic lensing and you you basically 3D print uh something that goes outside the head and then this can compensate for the shape and thickness of the skull and there's this kind of idea that you could actually create like a 3D pressure hologram inside the skull and so you wouldn't have a single Focus [41:14] you would have many fosi and you could like really you know get fancy with that um and that would be really fun to do um emerging field of sonog genetics which which is where we combine um you know virally introduced uh mechano receptors like the trp channels or there's a bacterial mechanical receptor called mscl um there's mech4 which is found in cgans and so you could introduce this is this is not fully noninvasive so there's a little bit of a downside but you could introduce a virus into a particular brain Target have it express the mechano receptor and then activate it with [41:54] ultrasound um so that's something we're looking into people are trying to do like visual prostheses with this and you know creating patterns of activity in the visual cortex it's pretty science fictiony kind of stuff um in terms of commercialization I think you know we're seeing more and more every year more companies coming out um in all the different you know in clinical Arena there's already you know the Insite um device which many medical centers are using and running a service to treat essential tremor um there are research systems that are uh pretty much TurnKey [42:35] so if you want to do mouse research with ultrasound um you know fous instruments will will build one of these for $50,000 um and they also have an MR compatible uh system for 100,000 that will work with a um I sound like I'm selling this I have no relationship with this company um but they have a NMR compatible system that'll work with a 9.4t system um and then some people are are working on wearable um ultrasound uh transducers and you know it's all very interesting I I think that we really need to get to something that's probably wearable but also with [43:17] multiple um transducers so that we have more control um over what's actually happening so that's it um and I'm happy to take questions more [Applause] questions I have a question oh yeah so with the slide with the Hemi neglect I'm just um kind of confused because if you're targeting the basil ganglia um and if there's a parietal dysfunction where hem neglect would happen how does that correlate um right so attention spatial attention is controlled by many structures in the brain um the [43:58] prefrontal cortex the prial cortex the basal ganglia and parts of the brain stem all all control so there's you know there's a there's a nucleus in the brain stem called The Superior calculus which controls eye movements um but it also controls attention um so so spatial attention is one of these functions that's so important that a lot of your brain is doing it um including uh the cod8 nucleus and the cod8 is also involved in controlling eye movements so both overt and covert attention uh how did you decide on the [44:39] frequency duty cycle uh parameters for your alround at the the private experiments really good question I mean we just we haven't really explored that space and we need to um it's just what sort of happened to work um and and it worked and it's like did you try a lot of things that didn't first or we we actually know um but there's there's no I'm not saying that any of this has been optimized so there's there's still like I said there's a long road ahead of us um to actually optimize um all of those parameters uh there is this idea that certain frequencies um cause excitation and [45:22] other frequencies and pressure regimes cause inhibition and so we can sort of you know change the excitatory inhibitory balance um by manipulating pressure and frequency uh but again early days and all that really needs to be expl I saw some of the stuff from the Korean lab that said that from from Korean studies that said that they that they knew which were excitatory versus inhibitory you play stock in that well you know it's it's an interesting hypothesis it needs to be tested more carefully okay oh yeah um I saw a little bit ago that [46:02] ultr was used to zap addictions away which seems very promising use case would it be the same methodology or was that just like P sciencefiction I haven't seen that study we um we tried to get a study funded where we were going to um look at ultrasound uh targeted to the you know certain reward centers as a treatment for alcoholism um so chronic alcoholism is extremely refractory um as you know if you have a relative or something you probably know about this um so you [46:45] know there there there was some interest but we never really got that study funded uh the the idea was it's a it's a structure called the ventral straum or the nucleus cumbent and it gets a lot of dopamine um and you know people think it's involved in um in addiction and in alcoholism so we're going to try that um never got funded so we never did the study but I'd be interested in the sub stud yeah so this is more about decentralized science and kind so it seems like you have a massive uh search space that you could explore exactly uh for this and presumably resources I [47:26] don't know how many people you have in your lab but um and then simultaneously there are probably many other labs that have trans similar transducers and access to similar you know um animals that they can experiment with um what is the feasibility of parallelizing a lot of these experiment because even if you did have the resources you you would be you would be limited by this by the sequential nature of like working in a single lab in a single place human beings um and so how realistically do you see any possib like um any possibility of parallelization of [48:06] these experiments uh and aeration of their results yeah so I'm going to say something Politically Incorrect not in this country because that's just not the way the NIH fund sence um if someone who wasn't right it might have in China oh okay you um because they do things big scale um yeah it's uh no I mean what you want is something like the Allen Institute but Allen Institute is focused entirely on on rodents but you know if we had an allen Institute for human Neuroscience I mean [48:47] I think they do a little bit of human Neuroscience there as well but that's the scale that you kind of need and that's you know you're talking about you know um I mean the entire NIH budget is about $50 billion and you know to really do this at scale yeah you're talking about you know a billion or so so um and then what are the I try quickly what are like commercially like how much money would there be to make assuming this were Technologies be successful yeah that's a question like how much like would it be more than billion I guess would it be more than T well no I mean [49:28] you can build these systems you can build uh you know a colleague of mine is building a phased array system he's getting components from China and it's about on the order of 50,000 per system um so if you wanted to have like tens of these systems you know say 50,000 time 10 or times 100 you know so it's you know it's within the um you know few Mill few million basically you could have a pretty robust Enterprise and start doing these studies so a lot there's a part the cap of D is like super biof focused and particularly around this and um there's a lot of [50:09] focus around um intellectual property and of these and so I'm I'm wondering if you had a bunch of scientists from all these different because I know you're saying it can't be done by the NIH like I completely agree that's one of the limitations that Michael was talking about earlier but suppose for example that um we focus bit on the supply side and we just have the scientist do these things anyways for free or whatever but we kept the tracker attract their contributions to um their this project and like uh basically allocated them a portion of the inal property that was created out of it do you think that it's realistic that um because then again that Inu property would have value you [50:50] know probably very valuable maybe um it could be reward all the scientists who parti do do you think that that is possible in any way uh it's possible you got some legal hurdles because uh the two institutions um that are involved here are the universities and the NIH and both of them have intellectual property interests that you would have to deal with and so Columbia owns technically they own all of my intellectual property and if I want to license something or patent something have to go through their lawyers um that's that's a real issue I can't really do things on my own [51:33] unless I sever that relationship I think I don't know if people try to get around that I I haven't done that I mean I have in the past you know licens some products through them but it's you know it's an issue as soon as you license something ni wants to know if that was supported by a grant and then you have to go through their patent process and so forth and and you know these are limitations but um you know universities have a lot of uh a lot of power and you know and a lot of legal um I'm not going to say too much about [52:14] tce that's why you need like Dows to uh like fund initial research with like grants so that people can have like the amount to live but then also go the IP that would go into the grant that would go into the Dow instead of having the IP go into like the IP licensing department of the University right right but you know changing that model is going to require you know a lot um I mean I think institutes that are like fully self-contained like the Allen Institute you know where you just have a you know huge endowment um but they also get ni money [52:54] and they also have University partnership so um you know you have to deal with that at some level Yeah you mentioned before that you are targeting viruses so here a practical experience and people have a chronic bbpb which is a recuring Leo they are trying to heal it with you know vable treatments I had this chronic Leo for one hour every doctor told me you're going to have it for the rest of your life I did a lot of resarch study from that and I find out when you have the basically viruses are bunkered in the Vesti ner and you have to go after the viruses not just m already go with anti [53:34] vetical drugs they give you you know volume they give you Clum that is going to improve vable treatment with ever going to help that the crystals which get release there they're going to vanish on but the problem is that in the rest the viruses are curing so you actually Target the viruses in order to get rid of the vertical so now your application is perfect for that you can actually Target the Val system n to see what is happening and then you can use like aob Dr to go after it you know that's an interesting idea that's a very anatomically it's a very complicated structure because you've got the you [54:15] know it's called the Bony Labyrinth and it's embedded in in the base of the skull basically um so targeting ultrasound would be really pretty tricky there but I I see what you're saying I mean you can have vertigo from acoustic neuromas from viruses from the the little ear Stones the otoconia get displaced sometimes um and there's sort of a maneuver you can do to uh get back into place and things like that um I mean ultrasound could potentially play a role there but um it it's a very an uh complicated anatomical structure so it' be kind of of difficult it's easier to [54:56] go things after things in the center of the brain you know because it's just a big melon question so for like the micro like uh why five Micron uh five microns is roughly the size of the capillaries so you have a a micro going through and it's like just fitting through the capillary and then you make it expand a little bit and then you can you know get the capillary to to you know loosen up a little so uh could you potentially like make it smaller or oh yeah yeah um I mean they make these micro bubbles in [55:36] all sizes they're originally developed I think as uh contrast agents for heart Imaging so when you get an echo cardiogram and they want to do a flow study or something they can use these microbes so they come in all different sizes we actually make them in different sizes and then we use a a filter we call a sizer but it's a filter to uh select the five micr size um thinking about the like wearables and home use how big is the I guess like margin of error for where you would put the device like if you have it off a little bit will you still get the results that you want well that's why we it would have to be packaged with a navigation system you know so you could [56:19] sort of move it around and see in real time you know you'd have a a computer that would show you in real time where the Focus was and then you could kind of like you know scrap it down or whatever um uh I think in terms of the tolerance it depends on you know what you're going after what the structure is if you're going after say like um a Target like for um uh you know the the anterior singlet um a particular part of the anterior singul is now a target for mood disorders and uh OCD and it's a fairly large structure so there's a good margin there but if you're trying to Target the um you know the acous the uh you know [57:01] the eighth nerve that's that's you know that's a different issue Al together all right um why do the lower frequencies penetrate penetrate deeper than the higher frequencies uh it's just the physics um it's the same with light low frequency light goes deeper than than high frequency when you're doing microscopy um it just has to do with the way it's absorbed and the you know the mechanical compliance of the tissue and things like that um it's it's a physics problem and I'm not a good physicist um does age play a factor in [57:44] this uh yes age is a very important factor and something that has to be looked at um and you know animal studies are almost always done on young health healthy animals but if we're going to apply this you know to geriatric um populations we we need to do and that's a that's a scarce and precious resource um you know animals because it costs money right to to feed and house the animals and so the older they are the more expensive they are or the more that they they've cost you so there aren't many places with geriatric populations of primates for example there are only a few places in the unit [58:25] States but I think it's really important to do those studies before we um you know develop this for for you know aging humans of which I'm one so we're all aging just different excellent all right is there any more any more questions well thank you so much let's give another round of applause amazing okay so next up we have uh Sage and um little bio on Sage is like a couple years ago I was uh I was doing the PHD in computer science so I guess I was doing a lot of Googling on [59:06] YouTube and just in general and Google knew that I was interested in Ai and uh SJ would always come up on like the front page of my YouTube channel and uh he has over 750,000 YouTube subscribers and literally a plethora of like any topic you could want like whether it be uh word to back glove uh Transformers like any sort of topic that you could think of he has like he's the modern day Donald can with the art of computer science uh for YouTube he's and um and now you're working on uh sports betting AI sports betting [59:46] AI yeah and um so I mean I'm just he's a he's a great um uh digital YouTube AI educator and um you guys should all definitely check him out on YouTube he hasn't he really has an amazing channel so I'm very excited to hear his your talk on um Ai and Longevity and the sort of science experiments that he has been conducting and I I think you're also going to talk about AI intersection of science which I think is like probably like one of the frontiers of science right now because you know all the sort of research assistants um they could probably be replaced by AI in many anyways and uh like we're just sort of [1:00:26] scratching the surface on what does it mean for an AI to be a scientist so without further Ado take it away AB right thank you Michael thank you everybody for being here I know it's super late but I'm going to try to keep everybody awake for the next 15 minutes so stay with me we have a lot of information to cover in this 15minute talk I'm going to try to keep it at 15 minutes we're going to first go over what is AI for longevity very important topic one of the hottest topics of the next decade then I'm going to go over my own journey in AI for longevity and at the end we together are going to build our own AI or longevity drug in 10 lines of python code all right ready ready for this all right let's get started [1:01:08] so what's the problem here everybody's dying left and right right so the top 10 causes of death are all related to different systems in the body whether it's the digestive system whether it's the heart whether it's the lungs whether it's the brain so you might be thinking that these are all disperate systems that they not interconnected that they have nothing to do with each other but the reality is that all of these diseases are interl and that they are all Downstream of the root cause of all disease which is aging and that statement has gotten me laughed out of Hospitals and Clinics and different Healthcare facilities when I say it because this is a multi-billion dollar industry and I'm sure Vincent knows what I'm talking about but this is something [1:01:49] that people's livelihoods is against they're paid to not say that aging is a disease but in fact Alzheimer's is a disease and you should perpetually take this drug to cure yourself of Alzheimer's forever so that me and my shareholders make money like everything else in this country but the reality is that aging is the root cause of all disease and as radical a statement as that might seem there are three different sources that I was inspired by to come to this conclusion the first one is a book that I highly recommend you check out it's called the gene and it's by Dr sart muker and I got the privilege of working with Dr Sid in La at the LA Children's Hospital Hospital on a bunch of genomics data and uh also for the US Veterans Administration and I got to see [1:02:31] firsthand how incredible a time we live in uh it is right now because we are seeing an explosion of omix data omix stands for molecular biology data that includes Gene data so you know what encodes who you are transcript omix data all these different omix we'll get into that later on but uh it's an incredible time to be alive to see all this explosion in molecular biology data data that did not exist more than a decade ago now exists and the second book is called lifespan why we age and why we don't have to by Dr David Sinclair actually people might know this guy raise your hand if you know Dr David Sinclair yeah he's he's he's gone on social media a lot but this is a great book he also has a podcast highly [1:03:11] recommend it um this Sinclair lab for genetics they've done Decades of research if you remember from the early 2000s there was this craze around you know red wine and reservatrol that came from their lab as well but all that antioxidant stuff it wasn't as impressive as what they later f found which is what I'm going to talk about the third uh I guess resource is my friend Brian Johnson who's been my friend for you know a few years now we used to hang out in Santa Monica and talk about Colonel which was his startup back in the day which was a brain interface startup and he was super into that um but nowadays you know I think everybody here has heard of Brian now he's working on longevity full-time and Brian and I have two different trains of thought here which I'm also excited to go into and what longevity means for you [1:03:51] so that's kind of my background so before before you and I get into building this in 10 lines of python uh we need to have some context here of how are we going to do this how are we going to build an anti-aging drug in Python in 10 lines in Python sounds ridiculous until you consider the context in that we can consider everything information for a second just bear with me for a second it's a little hypothetical experiment let's consider everything information for a second well the most fundamental unit of information then is the atom and what do we use with the atom we build things we build this desk we build this computer we build cars trains planes all these with the atom um and one of the things that we built with the atom was the computer and the computer gave us a secondary unit of information the bite the bite is [1:04:32] represented as a one or a zero right there actually happens to be a tertiary unit of information that people have been sleeping on and that is the gene it's not encoded as ones and zeros it's encoded as A's C's T's and G's and gez who here has had your DNA sequence before by show of hands just out of curiosity it's actually a large number of people given the privacy issues which we're also going to go into um I've had my genome sequence and it's literally a txt file on your computer you can double click on it and it literally just looks like AC literally everything about you is encoded in these numbers and so that's just something to think about okay so biology is information physical matter [1:05:15] is information everything is information okay so what evidence do I have here other than these books that I've read and these people that I've talked to that aging is a disease and that it's treatable well two years ago out of Spain there was this gene therapy uh that was applied to lung cancer patients it was a gene therapy specifically for lung cancer but as a side effect some of the patients actually experienced age reversal they experienced repigmentation of the hair their wrinkles went away so different body parts ex differentiated is the actual term for this right where ex differentiation is a term in the longevity community that means that different systems of the body lose their flavor or their distinction over time [1:05:57] which causes aging whether it's in the you know theal cells or whether it's in you know blood cells whatever it is and these these patients experience a reversal of it as a side effect they did not expect that but often in science the most crucial discoveries are accidents right Watson and Crick that's how they discovered DNA and so my point here is that this is a wide open space and if people are reversing aging by accident with Gene therapies and there's something about Gene therapies that we should look at now I am a computer scientist well I actually dropped out of Columbia three years ago when I was studying computer science as every Silicon Valley person should do right and what I came to terms with after dropping out of Columbia I didn't really [1:06:37] have an interest in biology when I was a student what I came to terms with after is we should use all of this computer science knowledge to figure out ourselves and how we work and that's what this explosion of biological data is going to allow us to do as people like myself who don't have any business being in biology I don't have a degreee or PhD but I will affect biology and I think you can too one of the ways that we can affect biology is by creating Gene therapies Gene therapies are essentially bits of information remember all this is information designed on a computer you and I can design this in python or a different programming language this information once we design it is then printed to a DNA sequence another way of thinking about this is [1:07:18] that it is uh it is a you know just like you would a 3D printer with pla or something plastic you can print out DNA material and U once this DNA is printed it is put into a transportation vehicle like an aav or some sort of virro vector or something and then you could take it as a pill or as an injection or something like that now if we just have a laptop and we want to you know hack on something we don't have access to all this molecular material we don't have access to a PCR machine or any of that and that's where Cloud bioinformatics comes into play there are companies like twist biosciences that allow you as a developer to then push your code to the cloud and print out DNA that they will then send to you for you to test out on different organisms with which we'll [1:07:59] talk about as well so where do we begin in biology this space is super complicated and as New Yorkers one way that we can look at data and understand it is in terms of a metro map what you're looking at here is a metabolic metro map it is a map of all the different path Pathways in a system of the body known as the metabone and we'll talk about all the different hes in a second um but as you can see nobody should understand this from you know right off the bat this is super complicated and you can see all these different processes happening here you know P glycer rates going up to the glycerol and the glycogenesis going downwards and all this stuff um this is super complicated and entire researchers [1:08:41] who I respect and admire dedicate their entire lives to one small part of this Subway M so it's super complicated but what we have learned in the past decade is that there's three Pathways in this metabolome that are super useful for longevity those pathways are bear with me for a second number one mtor which stands for Maman Target of rapamycin we'll go into that number two is cin which stands for a set of a subset of proteins dedicated to longevity and number three is amk which sounds sounds like a punk band but actually it stands for adenine monophosphate Kines which is another longevity pathway in the in the [1:09:21] metabolic pathway map now why are these three important well Dr David Sinclair in his book why we age and what why we don't have to in his lab you found that these three longevity Pathways when activated actually reverse and stop Aging in mice and in humans through these methods those methods include caloric restriction exercise reservatrol some of you might also know about cold water therapy ice showers all this stuff what's the relationship between all of these different seemingly unrelated activities they all cause stress in the body it goes back to Xeno Oris that which doesn't kill you makes you stronger and all of these things activate amk so every time your body is [1:10:03] under stress it's like you're telling it you need to survive you need to actually bunker down take what little energy sources that you have and have it last over time and um ampk is also related to mtor which is also activated when you take too much protein and protein actually increases your muscle mass so when you're lifting you actually feel great when you have a lot of testosterone you feel great but actually that's burning the candle faster so you're got you're actually going to die faster I know that sounds kind of counterintuitive but it's it's kind of like burning a candle too fast and so these things are all interrelated now these are just three Pathways in that metabolic map and they're just the first three that we've discovered um another [1:10:45] drug that is great for targeting ampk is metformin this is an FDA approved drug it was actually approved for diabetes and as a side effect helps people live longer um so there's something with insulin as well so there's a lot of uh discoveries to be made in this field now to me let's get to me for a second so what have I done here I am just a mere mortal again I don't have a PhD I'm your College Dropout I'm your YouTuber but I happen to reverse my age eight years in 11 months which actually beat my friend Brian Johnson in the Olympics by the way don't tell him that he's not going to like that but the way that I did this is not by creating a list of supplements to sell on the internet um I'm selling a sports betting bom that's [1:11:26] what I'm sing um the way I did that is just a simple tried andrue method that has been working for decades diet exercise sleep and healthy relationship SL mental health there's really nothing complicated about that but what taking blood tests three times in 11 months did do for me that I measured on inside tracker but you could do this on any website I'm I'm not I'm not saying inside tracker is the one you should use what it did do for me is it showed me what I was doing and if it was good or bad like for example just you know sleep more don't eat out so many times like you know a lot of sugar or you know don't eat burgers and fries every day like simple stuff I really don't think this is too complicated for people like [1:12:07] this has been going on for hundreds of years like there's a study called the Blue zones right and all of these places Okinawa Japan Sardinia Italy Linda California NOOA Costa Rica they have nothing to do with each other other than the fact that they all live past 100 regularly and their diets are totally different right youve got fruits and vegetables you got fish you got meat you got red meat you got Dairy everything that people say that they you know not into these people have been eating forever and they live longer and they're happier than all of us and so I think the reality of what we as near mortals can do to live longer is to a take blood tests regularly and B just try to eat Whole Foods you know regular moderation diets um and this is just kind of a [1:12:48] snapshot at my data at you know some of the different of the 42 different biom markers that they measured calcium sex hormone binding globulin alamin levels it's a liver protein and what it does is it measures that level um against a a threshold that is you know somebody of a given age and it tells me the Delta like oh you're your levels are plus 1.2 years older than somebody of say 18 which is kind of the Baseline and what you can do you can change your nutrients and mineral levels in your diet to try to uh make that better that number better and look we're all going to die this is not going to fix anything diet exercise sleep that's just a reality but it's a fun Challenge and it's a better way it's [1:13:28] a more holistic preventative way of thinking about health and wellness that being said I do want to have a side note here and that is to be weary because the US federal government is going to spend more on Healthcare than the military in the coming years because everybody's going to get old right and so we have an aging population just like Japan if you go to Tokyo now you can see exactly what we're going to be which is that a bunch of old people and uh that means that there's going to be a huge industry to sell longevity drugs to you so I just want you to be wary I don't actually have a longevity drug to sell to you just yet but I want you to be weary that's all I'm say here okay um so what can we do right this goes to building the thing in 10 lines of python code [1:14:09] what can we do if we don't if we don't like have a PhD and we don't work at one of these companies well you we could get a job that's always an option and there are 20 biotech companies that are hiring right now uh for people data scientist bioinformation people into space um new limit is one I was talking with one of the people here earlier about my friend Brian Armstrong who started coinbase now he's working on that this is a really cool one I would highly recommend new limit you know Google started Calico 10 years ago you've got Jeff Bezos spending all his money on you know aging very billionaire thing to do these things right so that's one option but if you're like me and you don't want want to necessarily go work for the man and you want to start something of yourself [1:14:49] let's let's talk about what that looks like so what I have here is a is a is a small little video of what's happening in the cell and it's actually not going to play because I'm not connected to the Wi-Fi but I'll explain connect we connect okay [1:15:22] one okay so basically it's not going to play it's okay no no no it's not the Wi-Fi I just I I did something wrong it's not you uh describe it I'm going to describe it so if you go to the smallest scale if you go to YouTube and you type in biological Machinery first video what's happening at the smallest scales when you really zoom in using equipment that was created in the past 30 years is it's a bunch of robots and they're going on these levers and pulleys and all these complex physical Machinery all this is happening at this molecular scale in a Pico second hundreds of thousands of [1:16:02] petabytes of data are being generated scales that you and I could not even dream of being able to fit into our threedimensional head this is multi-dimensional data hundreds of thousands of Dimensions so all I'm trying to say is that it's super complicated that being said as complicated as it is it is still information and like all information it can be organized and I'm going to try to organize it as best as I've understood it while working on it for you today one way to think about life at the smallest scales is a stack of layers just like a software stack in a computer at the bottom of the stack you have the genome the DNA this is what has the storage the hard drive all the [1:16:43] information about you stored in the genome now every disc reader needs every dis needs a reader some way to read the disc that's what the epome is and that's where you're seeing a lot of of the Innovation today data that's coming from the epigenome tells the body how to turn on and off all of these switches in the genome all the A's C's T's and G's when and when when and where to turn them on and off now there's also another layer on top of that it's called the transcriptome what the transcriptome does is it translates all of those onoff switches to different sections of the body and all of those translations need to be turned into applications there needs to be an app store for the body and that app [1:17:23] store is called the proteum proteins are the apps in the body proteins there are thousands of proteins and each of them does an individual task really well and one way to monitor all of the actions of these proteins is with a process monitor just like we would on a computer what are all the processes happening and that process monitor is called the Metabo and that's what that's that metabolic subway map that we saw earlier that's called metabol which monitors the actions of all these applications the proteins that are created through the this translation process by the transcripton that comes from the onoff switches in the epigenome that comes from the hard disk of all biological beings which is their DNA now [1:18:04] that we understand the stack of how biology Works how can we change it now bear with me for a second so DNA is a sequence of A's C's T's and G's right and in Compu computer science we learned that computers operate on sequences of ones and zeros so one thing is certain here well two things both computers and biology are both computers right that's for certain the other thing is that they both operate on sequences so there's some similarity there what if we could convert the sequences on a computer into the sequences of biology what would that look like well what it would look like is a map of gates of [1:18:47] logic gates of conditionals if if then statements if this happens do this and then do this and then go back and do this how do we do that well we're going to design a logic gate and the way that I we we're going to design this is a library called cello it's called cello on GitHub and I'll talk about it more in a second but cello allows for genetic circuit design in Python and once we design this and there's there's a couple examples this is one of them once we design this gate um and we're going to you know basically design it to create a new type of small molecule drug that is going to pull from a database of what are called lians and lians are sequences [1:19:28] of molecules that they come from different drug databases and we're going to tell it what machine to run on as a parameter the CPU number we're going to run it on our CPU because it only needs a CPU really to generate all these different uh variations of the base liend we're going to set these four different parameters the charge multiplicity solvent and then where to save it now these three parameters can be changed we can play around with them and we can use an optimiz ation strategy namely drum roll please AI we can use AI That's where AI comes in remember this data is way too complex for you and I to understand but that's where AI comes in and so AI is going to help us understand which parameters to use to generate the optimal molecule given this liand [1:20:11] database and this is all going to construct uh a molecule using a series of logic gates that are written in this cello programming language now this should be confusing which is why I have one more example from a different angle to give you a different view of it and a more holistic view of what we're trying to do here now I said that cello was one library but there's two other libraries that I want you to know about the first one is called coral and that's even more recent the second one is called biopython actually Great Name by the way they should have uh that's a great name so what we can do with these libraries I talked about cello another one is coral is once we import that Library we can create what's called a primer and the first thing that people do in a [1:20:51] molecular biology La is create what's called a primer and what a primer can do is it can you know inhibit or say amplify a certain Gene and in this case it's going to amplify this Gene basically it's going to make it uh so that it's going to be activated in whatever organism that we're going to insert this into and it's going to read DNA from a file that was uh sequenced from some organism maybe a human and then it's going to show you the sequence of a c's T's and G it's the code of biology right and so once we have that we can run a design process using coral on that data to create a bunch of variations and we can just test them out and right now um the process of testing it out takes a decade right the drug design process we [1:21:33] have clinical trials you know phase one two three and it's a huge huge uh choker on the drug Discovery process it slows it down way more than it needs to which is why Quantum Computing is very exciting and Richard finan said in 1965 that the way we're going to simulate quantum mechanics is through quantum Computing and imagine if we could simulate this on a biological organism that lived on a digital machine essentially simulating a drug trial on a computer in silico avoiding the need for the entire clinical trial process and being able to print a drug from idea to patient in you know n of one uh a day that's the future that I foresee with [1:22:15] this technology but we need a lot more attention on it for us to get there now let's say we have this algorithm you and I we we've done this we've got this 10 line code great where are we going to deploy this thing it turns out Amazon of all people they have a health omix platform when did when did this happen it happened two years ago under everybody's nose it's called AWS Health omix transform genomic transcript omic and other omix data into insights very cool stuff I really already want to get started with this don't you very cool stuff so that's one way that we could deploy this Amazon right that's one way that we could deploy this but there's other ways as well we could deploy to our own server we could create a paid [1:22:57] subscription service for omix data because of how valuable it is which brings me to me selling my genome as an nft which is kind of the you know gimmick for this talk right um this uh is actually a really important moment um in the history of web through because well first of all somebody bought my genome for one e and they resold it twice and they flipped my genome for three e so somebody made money and then made money and then made money again now if I would have uploaded my genome to any normal website forget blockchain you know blockchain's a Stam I'm not going to do that someone would have resold it and I wouldn't have made [1:23:38] it done but because it was encoded on this public transnational Ledger ethereum is blockchain that nobody controls every time it was resold guess who made a percentage cut of that pretty penny me and guess whose data it was in the first place mine so don't you think I deserve to get paid for that data and I think you do too the way to make that happen without one entity controlling it is through some tool that nobody controls 23 and me is the biggest American company that's where I sequence my genome that you know has the biggest database and it's already being sold to the highest bitter internationally right so we have a huge problem with Health [1:24:18] Data in this country and centralized Solutions are not cutting it right they're they're just not cutting it and so we need a way for people to first of all make money and one of the things that you and I all of us have that no robot or machine has is our biological data in fact it is our most valuable asset and we should make money from that we should not just give that away for free and the way to have a type of digital ownership where we don't have to become a surf to some server owner is to take that data and create a patent that lives across countries that patent is called an nft so while everybody else was Lo losing money with nfts biodata was making money so I think that this is [1:25:00] really the Black Swan for where blockchain meets AI meets science we're going to see this convergence of three different Technologies we're going to see genomics data we're going to see the blockchain to monetize this data for ambitious researchers who live outside the bounds of Academia who want to push forward the space who are willing to say no to the existing health status quo and want to create new innovations that those are the people who are going to make money from this and I look forward to helping them on that Journey um and I also want to say that Gene therapies that is designing information for living organisms with computers is not limited to just ending [1:25:42] disease we can create new types of abundant food we can create crops that have 100x yield um most of the fruits and vegetables that we eat today are already genetically modified right they didn't exist even 100 years ago human enhancement right and this is what we were talking about about the enhanced Olympics so that's already a topic clean energy um you know biofuel things like this and I also want to just end the talk on a very kind of out there idea to to end it with and that is plants are computers and if you've ever had a plant that might not have been apparent to you but I group plants I grouped tomatoes for a little bit and they were my little computers and I really love them because they were [1:26:25] able to take sunlight and convert that into energy and I did a little bit of research onto how they were able to do that and it turns out that they use a form of quantum superposition to perform photosynthesis and that means that that's a form of computing in a sense biology like we talked about is Computing and what if we could repurpose that Computing for not just you know individual a actions for regulation and disregulation inside of an organism but perhaps for everything else that we want to do with computers why do we have to have data Cent that look like the way they do what if data centers look like green houses that would much be much [1:27:07] more aligned with how the Earth is and in that way we are using AI not just for the longevity of humanity but for Earth itself so I think that synthetic biology is probably the most exciting topic um for the next decade it is where AI is headed and if you are interested um it's time to build um I'm a part of yes I'm a YouTuber and I I'm CEO of wager GPT and we're building sports betting AI uh but this is really my long-term goal and I and I and I a part of a venture capital firm as well mod VC I don't talk about it too much but you know I'm not very active but any excellent Founders in the space I [1:27:47] just hand over to them and uh we're going to get you funded if you got a great idea and I'm happy to do that and uh yeah thank you everybody any [Applause] questions how much did you win yesterday uh my members won a lot I actually didn't bet but they they were happy with the CH yeah so I guess what's interesting is you know the uh the pitch on gene therapy the analogy you make for the computer is like I guess you don't go right into the hardware to update you know find optimizations in in the iPhone so when you go back to the [1:28:28] omix tab like why focus on the team versus stuff that's more of the software higher level component that question this one yeah yeah like why focus on genomics versus the ones at the top no I I this is what I this is so much of the opportunity live right here I think genomics was really just the first data set that we got right you know found DNA but this these are new fields proteomic is like one of the hottest topics right now in molecular biology just from like a sideline looking in the subreddit and stuff like that you mentioned a couple of solutions um or like studies or whatever where [1:29:08] like oh this drug was meant to do this but it turns out it also people live longer um and do also in an earlier slide were like here look look at all these arrows all all these things are thought by aging so um think that some of those uh drugs were um capturing like a small part of the solution to aging uh and then like we kind of like we're we're only seeing like a part like a par view of it because we're trying to solve a different problem that's Downstream of the original problem like what do the inter play with that and like being Dev yeah I mean I have no idea but if I were to make a guess like based on what I've seen in AI so much of the progress [1:29:49] that's happened has been from taking away ideas rather than adding them for example attention is all youi was a SE paper 2017 right all these other recurrent networks all these different schemes to get data to reverse back into itself and you know none of it mattered you just remove all of that and attention which was just a mechanism deciding what parts of the input were most relevant that's all you need so I suspect something similar is in biology where it's just something of course it would have been right that's how turtles do it or this animal does it when you sold your nft did you sell the like how did that work did you put the NF did you put your genome on chain or [1:30:31] did you get the rights to it like how how did it work logistic wise that's a great question so like I said it's a.txt file AC's TS and G's what I did was I uploaded it to What's called the interplanetary file system ipfs um and what ipfs is is it's a database layer for blockchains and so you can basically think of ipfs as a blockchain but for storage when you upload something to ipfs it's not stored in one computer it's stored in a network of computers so your data is split into many parts and it's stored in bits and pieces across the world in what are called miners for ipfs and they're paid in what's called file coin but you don't have to interface with that it's it's kind of complicated but it's it's actually not [1:31:13] step one upload data to ipfs step two imprint nft on that data with ethereum and that process is just a transaction with the ERC 721 signature as all nfts have so but then anyone who had access anyone who went to openc could see that ipfs upload to your genomic data no so you had to unlock that nft by purchasing it and then you got to see the link oh okay got it got it and do do you worry about previous owners when when you sell it you could copy it but is that not a worry it you would think it's a worry but the [1:31:56] they would first have to buy it and if they decided to buy it and then they could technically find a way around it but there's schemes of basically blotting the data unless it has like an encoding like a key signature basically a hashing Alm I didn't get that far or you can Watermark it yeah basically yeah that's a better word for it yeah got it so okay and it's not a big of a deal if only like 10 people have it they're probably not going to or I mean maybe just owning that genomic like many people have owned the Mona Lisa but still owning the one that like like the [1:32:36] actual one is the most important thing versus like making photo copies of it yeah and then most importantly getting independent scientists to make money that's I think that's the primary motivation for me in this space that's what matters the most because it's really hard you know to get funding for your crazy scientific ideas and the more comfortable we make scientists with basically blockchains uh the better the easier it's going to be for them to get funding because there's a lot of money to be made here still especially for science it just it just makes a lot of sense to just get on board with this and if your project Works um your token's going to accr in value as well um actually there's a [1:33:17] great example of this like a legit example um out of coinbase it's called scub uh it's it's it's essentially like Brian Armstrong's uh science token project it's been going on for four years research up sorry research is where all the papers are uploaded they were a speaker here once yeah research up is a great example yeah how did you uh did you need a prescription or some sort of thing to get your blood drawn or how did that work logistically yeah so I just went to insid tracker.com and then I bought their premium plant it's way too expensive literally like um and it's [1:33:59] like 400 bucks and then I went to quest uh Nutri quest uh not nutrition but I think it's Quest Nutrition diagnos Diagnostics and then yeah I scheduled a blood test like a Walgreens and it sucked like I don't like taking blood tests right it's like a lot of blood and the lady like didn't care that's even worse and it's a lot too because it's 42 biomarkers because they got their premiums it's like four vials about that big um it's you know it's uncomfortable but it's also really exciting like you know nothing happened and then uh I got my data in five days after I got the blood test and it it was just basically a chart of all these things and honestly inside trackers like [1:34:41] whole AI thing is a is a joke I don't even use that it's totally wrong I just took the data downloaded it and then with python just visualized it different ways and it's just really interesting for me you could also probably just upload it to chat GPT and it would probably tell you exactly what a doctor would say exactly right I think you can do it for like 100 bucks just without the inside tracker part just go to Quest get a blood test and uh you can do a chat GPT exactly that's the most cost- effective way like how do I reverse my age given this yeah give me like the foods and diet exercise oh what do you do here okay here's what you do yeah exactly [1:35:24] yes um can you explain for the ins tracker the algorithm that was used to calculate the amount of years plus or minus that generated and I think you mentioned that it was based on standard of 18 years old how was that calculate does that an aggregation of data or what is that based on yeah so what they use they have a proprietary algorithm for computing that and it's you know they call it the inner age score and they recently updated it to to 2.0 and they updated that score between me taking this test and this test so it's actually unclear if it was even me or their score changed right but I don't like to say that too often I guess I'm saying that [1:36:04] now right so but my point in saying that is that is important is that it's proprietary so we can't actually know how accurate this is which goes back to show how important it is to have an open source algorithm that if you're going to take a blood test you have to know how it works and the most popular open- Source clock and there's several what are called Ang clocks that you can use to compute your Agee giv your biological data the most uh popular one is called the Horvath clock by Steve Horvath out of UCLA and yeah the orb CL is the main one and that's the one I would probably use yeah now when you do the Computing with the data and because we are using generative AI you know uh CPU is not [1:36:44] enough you need to now use GPU and for these are special platforms which where you can move the data from s to peles you can visualize you can store it and they have the very powerful platform with a big GPU on it I'm using it actually it's also yeah no I mean way this was just like a smaller example a serious need on G for sure or one of Sam alman's uh you know proprietary 77 trillion about application to and then do you think [1:37:24] there is someti [1:37:41] [Music] Sol besides yeah so I think first things first is we all need to be student of History right so I was working at polygon in 2021 which is you know top five cryptocurrency um at the time before FTX happened so I was working there as head of developer education and it was a lot of fun and I got a lot of insight but what but what we saw was a complete failure in the what was called the health to earn space you saw these startups that were saying like yeah you're going to earn this coin if you get healthy if you play this health game and the biggest example was sweatcoin it was called sweat coin and sweatcoin was [1:38:22] around for about 3 years and it sounds just exactly it is exactly how it sounds in that you earn sweatcoin if you run and then you could exchange sweatcoin for different gifts in a gift shop and actually you know sweatcoin did ACR in value and people were making some money um but then for whatever reason they just the tokenomics it wasn't sustainable which is almost always the case the tokenomics is not sustainable in the long run and I'm just waiting for the crew that's able to actually make it work now who did make it work openc kind of made it work openc is kind of this hybrid centralized decentralized model where there is a dedicated team and they are you know in one office but they don't own the U art the nfts that are [1:39:04] posted on their platform they don't own my genome but they help me list it and they were a discovery service for other decentralized buyers so I think that is something that is a signal a Marketplace for Health Data it's something that I've actually thought about for years but you see eight yeah or maybe but then none of them are super successful yeah I think it's really hard to make the tokenomics work for this no they don't have token just trying to to owner [1:39:50] Inc so it's to Sol [Music] there well they say that you know seasoned biotech investors like me they say that it takes like 20 years for a biotech startup to be successful that's a centralized it is it should it should be faster I I totally agree with that uh you know there's always one Golden Goose in the pile of C right like coinbase is our shining American example of a crypto [1:40:41] exchange that is reputable that you can trust with your money it's not going to collapse like FTX did uh and they're fully regulated right so uh and every everybody else was a scammer at the end of the day right FTX everybody else Gemini everybody else F so I think that's what's going to be the case with like bioh Health Data like there's going to be one company that does it Well if I knew the answer I'd be working on I don't know the answer yet [1:41:19] ipto yeah no I mean you know bitcoin's going up right now so people are coming back crypto as it as it always does the 10th cycle here we go on any other questions well thank you that was an amazing Applause uh fantastic fantastic um and yeah just enjoy uh thank you all for coming out we have the other one on February 22nd with the enhanced games president of the enhanced games Ain also fun fact about Aaron um I don't [1:42:02] know if any of you guys followed the Hulk Hogan Gawker situation with Peter teal but he was actually the one that came up with that idea to uh shut Gawker out of business he was the guy that actually pitched it to Peter teal to fund it to find Hulk Hogan to sue Gawker to get back at Gawker for outing Peter te's G so very roundabout but he came up with that idea so he's a very interesting person um and you guys are all welcome to come so uh thank you guys for showing me up show [Applause] [1:42:48] up cool [Music] yeah walk out and I have my [1:43:09] two so [Music] much [Music] [1:43:41] [Music] djbl DJ that's longevity [Music] [1:44:01] [Music] [Music] am [Music] how