Transcript

0:00:00 Intro Montage

[…]

0:01:04 Intro

I am the experiment. You can write your own story. Not stop trying and don't give into the fear.

0:01:09 Vahan Simonyan

Well, I mean, COVID itself — we know that it's an experimentation outcome of human experiment with an existing virus. Is that — is that accepted consensus across the scientific community? Most scientists agree that COVID is not a completely natural virus. There are some manipulations done to the virus, although I don't think that it was intentionally created or designed to be some kind of a weapon or a source of pandemic, because the difference between a mistake and a weapon is the ability to control and target it.

COVID was not controllable and not targeted. So I don't think an adversarial design was put into it. But the way it was designed — we know humans were working with this virus, trying to learn about the function of this virus, and somehow it got out. Imagine if people would try to design it with a bad outcome. It would be much, much harder, much, much worse.

Yeah. So... but I don't think — I don't think COVID per se was a biological weapon. I think it was a mistake that it ran out of a lab. But I mean, scientists do these types of experiments all the time. We design viruses. I mean, for God's sake, we had projects like we would design viruses to cure cancer — a virus that had pieces that would go on to target cancer cells and express molecules that would fight cancer, or a virus that would deliver CRISPR/Cas.

0:02:41 Justin McMillen

Was that what you call gain of function research?

0:02:42 Vahan Simonyan

Yeah. So there's gain of function, there is loss of function too, there are different types. But when people say "gain of function research," what does that mean? Okay. So there are a few types of reasons why humans get diseases — biological reasons. Sometimes cells that are supposed to produce something, they stop doing it well.

So you lose the function of manufacturing certain types of proteins to execute an important biological function in your body. Then, in this scenario, for example, with the virus — we can deliver the protein which the patient cannot manufacture anymore. And that would help the patient. I'll give you an example. A blindness example. Let's say there are specific proteins that are supposed to be produced in your eye, in the sensory layers of your eye, but with age or some genetic predispositions, people stop producing these proteins so they can go blind.

So we can take a virus. Let's say you take one of the deadliest viruses — polio virus, which is a deadly paralytic, neurotoxic virus. I can make it so that it's not toxic anymore, but it can still keep the function of that virus being able to infect neural tissue, but without making people sick. So that it's a good delivery machine.

Now I have this beautiful truck that I can load with anything I want. Why don't I load it with the gene that manufactures the protein that the human cannot? That's it. It targets the neural tissue. It goes and starts expressing that protein in that tissue, and the virus starts expressing mRNAs. But then that's perfect — because your body, your cell that is infected with that now non-dangerous, attenuated virus, sees the RNA and starts producing proteins from it.

Those proteins that your own body stopped manufacturing — that's a wonderful way of doing it. So, we have to be careful not to demonize virus research. In fact, it has a huge potential to help humankind — from cancer, from blindness, from any number of diseases. Today I'd say there are more than like 100 different diseases that are targeted with viral delivery.

But the fact that COVID got out of hand as an experimentation outcome and became this huge pandemic — and most of the scientific community — obviously, I'm expressing my own opinions, which I agree with. So I don't think COVID was intentionally designed to harm humans, but I think it was a lack of consistency in the procedures that are supposed to be in innovative labs.

0:05:32 Justin McMillen

That's what made it —

0:05:32 Vahan Simonyan

Okay. That's why I'm — I'm just curious, why would you not think that it was designed as a bioweapon? Oh, well, like I said — the control and the targeting. Imagine if you are trying to design a weapon and you don't want a weapon to go and blow up any time while you're still working with it, or you don't want it...

If you are — let's say, imagine an adversarial power who tries to be adversarial towards us. And they're designing a virus that can kill them as well as it can kill us. So there is no — that's not a weapon, that's just dangerous play. So COVID had none of the controls built into it. Like, if they were building it so that it could only affect one certain group of people, that's...

0:06:23 Justin McMillen

Now you're talking about targeting.

0:06:23 Vahan Simonyan

Okay. Yes. Control and targeting. Then you can make that thing go on when you need it, off when you don't need it. Targeting is different. If you want a virus to harm those who you target, but not those who you don't target. But that's a very dangerous field that humankind is now in. The reason is — like with CRISPR...

0:06:48 Justin McMillen

And I'll explain in a little bit what CRISPR is. But before we go further — because we've been rolling — oh yeah. Which I love all of this. Yeah, I'm already riveted by you, which is normal. You and I will talk for hours and hours and hours, you know, because I mean, you go anywhere.

0:07:06 Vahan Simonyan

How about...

0:07:06 Justin McMillen

How do you know all of this stuff? What's your background?

0:07:07 Vahan Simonyan

Yes. My background — I am a scientist. And my first education and background is actually quantum chemistry, a master's degree in chemistry. That means a lot of experiments, explosions, all that stuff. And then I did a PhD in quantum physics, and then I did a separate doctor of science degree.

So I have two doctorate degrees, again in nanotechnology and quantum physics. But then, about 30 years ago, after I designed some of the important aspects of a physics supercomputing center for physics computation, NIH stole me from the university. So I helped design some of the computing centers for scientific computations. Never a computer guy — always a scientist. But because I needed a good tool, I had to design a good tool for me. So that's how NIH saw value in me — NIH is the National Institutes of Health — they stole me from academia. So I built these capacities for them too. So I ended up being a senior scientist at the National Institutes of Health. I worked with the Human Genome Project and other types of biologically important subjects. And then, when genomics became the thing in the clinical space, FDA started getting the first submissions for large genomic data.

Believe it or not, the first genomic submission — it was for sofosbuvir, it's an HCV antiviral, an antiviral for HCV virus. The size of the files that FDA got was 13 times bigger than the entire data storage of FDA.

0:08:58 Justin McMillen

Wow.

0:08:58 Vahan Simonyan

So they panicked. And then they went to NIH just to help. And NIH said, we can't help, but we can lend you a hand. So a doctor — a Doctor Simonian.

0:08:58 Justin McMillen

Yes. That's me. Okay, okay. Keep going. Sorry.

0:08:58 Vahan Simonyan

So that's how I ended up at FDA. And then within six months, we built something for them. Yes. Within six months, we built the capacity to analyze genomic data. And then we moved to building an ecosystem — an entire ecosystem, because you recognize this was just the beginning.

And then all companies started doing genomic experiments, trying to target diseases by means of genomic design — viruses, proteins. So, does that mean that you would determine what the genes are related to a particular illness and then target your treatment based on understanding those genes?

0:09:47 Justin McMillen

As a scientist on a pharma or academia side, usually you would try to understand the mechanism of the disease.

0:09:47 Vahan Simonyan

Your body's like a huge chemical factory. It produces a lot of things. And imagine if you're manufacturing a car — there are places that make tires, places that make windows, places that make engines, and so on and so forth, and then you assemble it into the big thing. Except multiply that by two trillion. That's how complex your chemical factory is. So but then something goes wrong.

There might be something in a manufacturing line which is broken. First, we as a scientific community study what is broken. But then you cannot just go and fix that one thing because you have billions of copies of that broken thing in your cells, in your tissue. So what you would do — you would design a medication or another genetic machinery, maybe a virus with certain genes, maybe a protein, maybe some small molecule that floods your cells and fixes the broken parts.

So that's what it means to develop genetic-based drugs. You pretty much design the viruses to deliver the genes, or you can directly make a vaccine out of RNA — RNA vaccines, for example, for proteins. So this is what companies and academia are doing a lot of these studies. Once they figure out the reason for a disease, then we try to go and fix it.

In animals we reproduce that disease in the animal, and we try to fix it with drugs. After that's done, then you show it to FDA, the Food and Drug Administration, and you say, hey, look guys, I found a way and it works in animals. Then if FDA likes what you have done and it was scientifically rigorous enough, then FDA will come back and say, now you have to do it in humans.

So you find people who have no alternative to being cured from that disease — let's say it's a cancer patient or some kind of a terminal condition where every other alternative has already been tried. So those are good candidates for this new drug treatment. And you start dosing it specifically to understand what a safe dose is. Number one — your drug should be safe, because if it's not safe, it can kill the patient.

Even if it cures the disease, but it can kill the patient — so after you figure out the safety then you start doing efficacy. Which dose is still safe but efficacious? So that's how it works.

0:12:22 Justin McMillen

Yeah.

0:12:22 Vahan Simonyan

So as a result of all of these studies, companies generate a humongous amount of genetic information. So then FDA started requesting them to submit that information so we could validate. But they didn't have the capacity to analyze that information. That's when I got recruited — I was given a big team and we started designing an entire ecosystem. We call it HIVE — High Performance Integrated Virtual Environment. In this environment, not only were we capable of accepting the data, but also analyzing the data in an independent manner.

The difference between truth and opinion is — opinion is yours, I may have a different one. Truth usually is one, in the scientific world. So if you analyze data and come to a conclusion, and I analyze using my own means and come to the same conclusion — there's a good chance it is closer to the truth. But if you have your opinion and I recompute that data analysis and come to a different conclusion...

Now there is a problem. Because science has to be objective and truth has to be somewhat singular.

0:13:37 Justin McMillen

Yes.

0:13:38 Vahan Simonyan

And so FDA's capacity to reanalyze the data using its own means — that was very critical for our industry. So let me see if I understand this. So you have scientists doing experiments to try to get drugs approved, looking at things like efficacy and all of that. And then they come up with their data set, their set of information that says this is what we believe to be true.

0:14:00 Justin McMillen

Yep.

Now with HIVE — High Performance Integrated Virtual Environment — which is this system you built, which is an ecosystem for computation — that becomes a separate set of lenses, an analysis to validate or verify whether what the scientists said was true.

0:14:27 Vahan Simonyan

Absolutely. That's — you got it. You nailed it.

0:14:27 Justin McMillen

Okay.

0:14:27 Vahan Simonyan

It's just independent analysis of the same data set to come to the same or different conclusion, and to cross-validate what has been done — without any human biases, because different scientists with different biases are trying to prove their point.

So how is HIVE used for all — yeah. So HIVE is used — now HIVE is used in a variety of different analytics, more than 100 different types of analysis. HIVE is used inside of FDA, and there are hundreds of different ways you can use it outside. So the Center for Biologics Evaluation and Research — CBER — these are the guys at FDA who are doing genes, protein therapies, or anything related to biological origin of the drug or treatment methodology.

And then there is CDER — Center for Drug Evaluation and Research — which is also using HIVE. Those are smaller molecules. But now also some peptides and genes — the lines are kind of blurred between CBER and CDER now. CBER uses it, CDER uses it, CVM uses it — Center for Veterinary Medicine. CFSAN uses it — Center for Food Safety and Applied Nutrition.

Center for Veterinary Medicine — yeah, data for veterinary medicine. So there's a whole — you're also helping animals?

0:16:13 Justin McMillen

Yes.

0:16:13 Vahan Simonyan

Of course, because the safety of the food is important. And like antibiotics are important — which are going into our animals — and bacteria become resistant somehow. Somebody has to monitor if there is antibiotic-resistant bacteria in our food chain. Because once they jump to humans, now we're in trouble.

0:16:31 Justin McMillen

God forbid somebody's sick with a bacteria that is resistant to all antibiotics — there is no way you can cure that individual.

0:16:31 Vahan Simonyan

So there is the Center for Veterinary Medicine and they do a lot of stuff — this is just one example. Yes, they use it. And then also the Chief Scientist uses it. Center for Devices and Radiological Health uses it for diagnostic devices.

0:16:54 Justin McMillen

You know, in vitro diagnostics, in vivo diagnostics, all types of things.

0:16:54 Vahan Simonyan

Is it — okay. So this is a tool that's at the FDA. It's another way for them to understand whether or not things that are being presented are valid enough. So it's sort of a gate that has to be passed through to get to the public.

0:17:15 Justin McMillen

So does every drug have to use HIVE?

0:17:15 Vahan Simonyan

No, no. Okay. First of all, we have to understand — most of the drugs and products submitted to FDA are still not based on genetics or big data. This is solely for big data genetics. Complex imaging also — some very large, massive data analytics — some complex analysis, all that stuff goes to HIVE. But there's a lot of classical trajectories. You know, you give the drug to patients, you monitor their fever and their outcomes, and you do some basic statistics. You don't need that tool. You don't need a tractor to put a nail on the wall — a hammer is enough.

0:17:38 Justin McMillen

So HIVE is this massive machine with huge power, but you need to use it for appropriate reasons.

0:17:38 Vahan Simonyan

How many of the drugs that are coming out — or different therapeutics — are genetic? There are thousands of things which are genetic coming to FDA now. But I, because I left FDA a few years ago, I don't know the exact numbers. So I'm not going to make them up. But I'll say when I was there, the numbers were in the hundreds per year. Okay. So now it might be bigger.

The genetic industry is growing 30% every year. Biologics — when your drug is derived from biological tissue, proteins, or blood, or genetic is considered — peptides. So biologics — every year there's 30% more biologics coming. Drugs — I think it's stable. They're not growing anymore. There are still new drugs being submitted every year, but the number is not growing fast. But biologics every year — 30% more drugs being submitted every year.

0:18:30 Justin McMillen

So this is huge.

0:18:30 Vahan Simonyan

Yeah, that makes sense. It's probably more expensive to create these kinds...

0:19:01 Justin McMillen

It is.

0:19:01 Vahan Simonyan

So much more expensive. And it's much more difficult to validate — because biologics is very complicated. Some protein in your body may perform so many different functions. Like, how do you target it to the disease tissue but not to any other place? Or how do you make sure that it has the good effects but not the adverse ones? Or even if it does have an adverse effect — how do you control the adverse effect? It's much more complicated chemistry. So it's more expensive because of that. But the promise is huge. Huge.

0:19:24 Justin McMillen

Let me give you an analogy here.

0:19:24 Vahan Simonyan

Imagine if you own a crystal shop and a fly flies into your crystal shop and you definitely want to get rid of that nasty creature.

0:19:47 Justin McMillen

I'm picturing you running around.

0:19:47 Vahan Simonyan

Oh yeah. So you have a choice. You can take a hammer and start swishing it around. Eventually you might get the fly, but you'll break the whole store. But imagine if now we are approaching it genetically, in a more designed, targeted way. So instead of taking a hammer, I would release 100 ladybugs.

0:20:08 Justin McMillen

Ladybugs?

0:20:09 Vahan Simonyan

Larvae love to eat flies. Yes — the moment they catch them it's gone. And they don't break a single crystal. So that's the difference between targeted therapies and non-targeted therapies. Humira is not a targeted therapy — it's trying to get rid of the fly. So the same way, biologic drugs — we have more control to direct them, to target them.

And because of that, this industry, once it matures, will do less damage. And if you want, I can give you an example — like antibiotics and bacteria. Yes. Bacteria destroy human tissues — devastating effects if you didn't have antibiotics. But antibiotics are very dangerous because they're not targeted. If you drink an antibiotic, the number one thing it hits is not your disease tissue — it's the healthy biome in your gut. It starts destroying some of the important bacteria.

Yes. And that's why a lot of people, after the spike in use of antibiotics, had liver issues, blood issues — there are so many diseases, like C. diff. And of course, there are so many things. And also, bacteria learn to resist antibiotics. This is like a catch-22 situation. But the new type of therapy that people are working on these days, for example, is phage therapy. Phages are viruses — except they don't make us sick. Mostly they make bacteria sick.

0:21:25 Justin McMillen

Why don't we use the natural enemy — just like a ladybug against a fly — we can use phages, which are viruses of bacteria, to kill the bacteria.

0:21:25 Vahan Simonyan

And phages are so specific.

0:21:40 Justin McMillen

I was going to say — they only kill that one. So you could choose. You could be like, this is the bacteria that's creating the problem, and you could attack just that one. Because you can target the therapy.

0:21:41 Vahan Simonyan

The humans already have phages in their body — is that normal?

0:21:58 Justin McMillen

But phages are usually more characteristic of environments when you have a lot of bacteria. Of course you have bacteria in your gut — in fact, 50 times more genes are in your bacteria than in the entirety of the rest of your body. So yeah, assuming there are no phages in our bodies would be unreasonable. Of course there are phages.

0:21:58 Vahan Simonyan

Phages are being — so this is an example of genetic targeting of disease, instead of like a careless swishing of the hammer trying to kill the fly situation. Most simple molecule drugs are that way. But because it's very important — I do not want to discourage anybody from using medically-advised drugs. Drugs work. Because bacteria can do even more damage if you don't take them when you need them.

So today, whatever the doctor recommends — if this is the best choice, that's the best choice. People should use it. What I said should not be taken as an advertisement against existing drugs. This is a very important statement I want to make. But I was telling about the incoming future that we as scientists are trying to learn — can we design better drugs in order to target diseases?

Because cancer is a horrible disease. Bacterias are horrible diseases. If you don't take the drugs, even though the drugs may damage you — the alternative might be much worse if you don't. It can kill you if you don't. And then when FDA or other organizations look at this, we pretty much look not just at the benefit but the risk and benefit.

And then all drugs almost have risks. Yes. And so if the risks outweigh the benefits — that's when you have to be careful. And probably your doctor should advise you not to use it. But not me, I'm not a medical doctor. But if benefits are larger than risks — even having the risks, especially the ones that could be mitigated later with another drug — then it's advised to use the drug, even the toxic one.

Look at chemotherapy. Look at radiotherapy. They're all toxic for a human being. You know what the number one side effect for chemotherapy from cancer is? Cancer. Really. Yeah. But the chances of that happening are less than the cancer you're treating. So you see — this is, I don't want things to be taken out of context. When I advocate for biological drugs, I do not want it taken out of context and taken as medical advice.

0:24:08 Justin McMillen

Yeah. No, it's good that you said that. Yeah, it is. You have to put a disclaimer in there anyway. So you heard it for us. So I appreciate that. So — you, so therapeutics that are designed to target specific things more — the ladybug, not the hammer.

0:24:27 Vahan Simonyan

Yep.

0:24:27 Justin McMillen

And that is the emerging — 30% a year — this is growth.

0:24:52 Vahan Simonyan

That's right.

0:24:52 Justin McMillen

And now HIVE — you said you're not at FDA anymore. So do you — was HIVE yours or was it FDA's?

0:25:18 Vahan Simonyan

So I created the technology, but I gave FDA an unlimited license to use it — it's still being used. It's being used at FDA. Of course, they have a big team at FDA that is developing it further and supporting it.

And we are now — I'm part of the industry. We continuously develop and we actually give the new tools to FDA as well. Because this way, the industry loves this — it facilitates interactions with complex data sets outside and inside. Imagine this scenario. Imagine you are the CEO of a drug company, a biologics company, and you want to submit your data to FDA.

If you know how FDA is going to look at your data and analyze your data, you can do that before you submit to FDA. FDA would love that, because nobody likes when drugs get delayed because the analysis is not done right. Every person working there is a decent human being, a good scientist who wants to have the drugs in the market helping people. Every time the analysis is not done right...

There are questions and contingencies that take time and effort and millions and millions of dollars — and people die. And these people are not just numbers — somebody's child, somebody's mother. So FDA would love if companies use these types of tools — internal FDA tools — in order to analyze. So if there are discrepancies, those discrepancies are addressed earlier rather than later.

That's why HIVE has significant value outside of FDA too. You work with different biologics and drug and device companies. They can sort of test it before and they get a good sense — okay, this is good, we can go ahead and submit. An exact degree of probability that it's going to make it through. And everything's going to come out well. But this covers the scientific review part. Of course, there is a clinical part and other things — every drug has multiple aspects of review. So HIVE is covering the analytical data analysis part.

0:26:49 Justin McMillen

So... your trajectory. You get sort of pulled in, and you're getting pulled into computation, but you've got this quantum physics background. Yeah, that's helpful. Yeah. So now you're building stuff for NIH, then FDA, and you decide to move into the private sector. And I think the genetics piece is interesting too — because you had a short stint at... you were at CRISPR for some time?

0:27:13 Vahan Simonyan

Yes, it was at CRISPR Therapeutics, when CRISPR Therapeutics were establishing a platform — and I'll explain what that means in a bit — for gene editing drugs. So CRISPR Therapeutics is the very, very first in vivo gene editing company to release — actually, like Chevy. So what it is — let me explain. From bacteria we learned they have these biological scissors that can cut everything. Scissors. Biological scissors.

So this is part of the bacteria's instrumentation to fight against phages. But brilliant scientists — Doudna and Charpentier — figured out how it works, and humans were able to take that machinery and use it in a virus. Put it into a virus. And this virus, we can infect the disease tissue — it can start doing some benefits, like killing — like a gain-of-function type destruction of the bad cells.

So the good ones can be expressed, for example. Yes. So that's called CRISPR. I'm not going to go into the very long abbreviation — Clustered Regularly Interspaced Short Palindromic Repeats — okay. So CRISPR Therapeutics used that for β-thalassemia and sickle cell anemia — two diseases. Sometimes because of a genetic condition, platelets become misformed and misshapen.

They look like sickles — that's why the word sickle comes from. Yes. Sickle cell cells become like sickle-shaped, and they cannot flow easily through our micro-vessels. They get stuck. And then of course that may result in a number of diseases because you don't have blood flow — the tissues get undernourished and under-oxygenated. Inflammation is in place. It's a very painful disease.

But CRISPR Therapeutics scientists learned they can use the CRISPR machine — they can prohibit the sickle-shaped cells to be produced and activate the fetal-form cells. I'm simplifying right now for a bigger audience. Yes. So that drug is an amazing drug. People who have been in pain for decades and have a shorter lifespan — obviously for obvious reasons — under constant inflammation and pain...

So this drug is not that you give it to the human being and that's it — the virus takes it into the human and starts producing these CRISPR scissors. Imagine little scissors that only cut and destroy in one specific place, and that specific place prohibits sickle-shaped cells from forming. Instead, other cells express and people are getting better.

0:30:06 Justin McMillen

And that's when you were at CRISPR?

0:30:06 Vahan Simonyan

I was helping them by data analysis and — for when I talk about this for a second, because I think people don't — most of the vast majority of the population doesn't understand, not that they don't understand what you're saying, but just how unbelievably advanced we've gotten.

0:30:28 Justin McMillen

Yeah.

0:30:28 Vahan Simonyan

Humankind is amazing. Yeah. I mean, you're talking about using a bacteria that can go in and — and the thing that I want to get clarity on is — you're talking about editing genes. Okay. So what are genes and how would genes play a role in the sickle versus not? Can you — when you said yes — okay.

So let me explain what genes are. Again I'll use an analogy. Okay. So in your family, most probably you have a recipe book. You know, I have a recipe book in my family which is a mixture of Armenian, Thai, American, all that stuff. So your recipe book and mine may be slightly different from each other? Yes. So when you or your wife try to cook something, you read from a recipe book and take the right ingredients and mix them together.

And here comes the delicious dinner for you. Now imagine — take that to a factory. There's a recipe for how to make a car. Yes. A car factory. A lot of different parts. But then you make cars of this type, of that type, a different recipe. So that's how cells use genes as recipes. Recipes tell you — synthesize this protein. The moment it's done, it rolls into this shape, falls into that shape, it becomes an instrument.

And then one instrument attaches to the other, to the other, to the other — clubs into these machines — they become machines. The vehicles inside of your cells. Imagine, instead of the car factory worker gluing a piece of metal, welding it to another piece of metal, attaching a plastic — your cells are factories doing it based on a recipe, which is the gene.

Genes are recipes. Now imagine what would happen if — because of old age, or because of some other condition, or because of radiation from space — it comes and burns the page with half the recipe. And basically, if you're trying to make that recipe which is important because you love to eat the dinner — of course you don't know the recipe anymore.

So you'll make mistakes and you might cook something which is not of the right kind. So these are all genetically predisposed kinds of diseases — missing information. Missing information or an error in the wording. Yes. Okay. Like if you had "salt" and instead it says "soul" — how do you add a soul into the soup? You cannot add the soul in a soup.

Yes. But what I'm saying is — these are mutations, or deletions, or sometimes extra pages can get stuck inside. And you can make a mistake and mix the cookie recipe and the mushroom soup recipe. I'm not sure that's a good thing to have. So the same way, genes can make mistakes — all types of mistakes. And that's how genetic predispositions work.

That's how people get sick. And then, all we are doing — gene editing — let's say there is a recipe book with a mistake in it. But then you have like ten different books. And if this soup is not the right soup to cook, maybe the other one is okay. So what if we go and completely shred the bad recipe?

We destroy the copies of the books which have the bad recipe. Yes. So you are only left with a good recipe book? Yes. So if you keep cooking — and you don't have these shredded pages — who cares? I'll cook something else. And then in the human body there are redundancies. Once you destroy the one which is not working well, your body sometimes starts using the good version of it.

0:34:06 Justin McMillen

So these kinds of drugs that do this — like that drug for sickle cell anemia — probably takes a while to work well, over time. It's like your body's got to go through this process?

0:34:26 Vahan Simonyan

Yes, yes, it is right.

Depends on the age, depends on the ability of the body to recover. Age is an important factor. Also how long they've suffered. Because it's not just — even if you remove the reason for the disease, sometimes the body has taken such a toll already. That's a long process to recover. Sure. And also it's never 100%. Think about this — imagine if there are ten people in a room and you want to help everybody, give a good recipe to them.

Sure, no problem. Now imagine if you have 100 billion people in a room. And then you're given a bunch of books with the recipe — so most probably not enough for all 100 billion. Also, how do you know which ones' recipes are broken, to give it to them and not to others? Right. You'd have to get that information. And 100 billion is the number of cells — the hundreds of billions number of cells — in the human body.

So for these kinds of therapeutics that are gene-informed, there's a gathering of data from the patient through testing.

0:35:32 Justin McMillen

Yeah. It's very important.

Yep. And then — I guess my next question is — so these therapeutics... this isn't the right way to say it?

0:36:05 Vahan Simonyan

Yeah. Therapeutic is the more general way, instead of saying biologics or drugs.

0:36:05 Justin McMillen

Biologics or therapeutics. So they're looking for — when they're studying them and trying to bring these drugs out — they're looking for a specific... not common, but like a set of genetic factors or mutations that exist in some portion of the population. How much... is there a future where we would gather genetic information from one person, and even though they may not fit into a category with a bunch of other people, that we could produce drugs that are completely personalized for that one person?

0:36:31 Vahan Simonyan

Correct. There is that. Yes, yes, yes. There are multiple intermediate states to it. There are a lot of genetic similarities in human beings — we are very, very much alike. Yes. Less than like a percent of our genes are different from each other. In fact, much less than 1% of the genes.

So, 99% — to say 99.8% I think is the last number — we are identical.

0:36:59 Justin McMillen

1 in 1,000 different cells fighting all the time. Oh yes. Oh yes. That's human nature.

0:36:59 Vahan Simonyan

This is also in genes. But so — yes. Every individual has a unique genetic makeup. Having said that — unique in 3 billion letter textbooks that we are made of, the genome. But let's say if I take this particular recipe, yours and mine might be exactly identical. Humans have 26,000 genes which can, after gluing and cutting and pasting, be turned into 400,000 RNA transcripts, we call them. So out of 400,000 different ways you can cook proteins in your body, that particular recipe might be identical. Think about 400,000 RNAs in humans and 8 billion humans.

You've got to have a bunch of people with the exact same recipe, despite all of the differences between us in other recipes, other genes. Yes. So if we make a drug which specifically works for this version of the gene, there's a good chance the drug will work for a bunch of other people also, if that's the gene that is broken. If we are fixing that gene or its function, then there's a chance — you divide 8 billion by 400,000...

0:38:00 Justin McMillen

Oh, nice.

0:38:17 Vahan Simonyan

There's a good chance it's going to help other people as well. That makes you see — it's not really truly personalized in this sense.

0:38:17 Justin McMillen

I was going to ask you — is this personalized medicine?

0:38:17 Vahan Simonyan

Yeah, it is personalized medicine. It's called personalized medicine — not because it's just for one individual. It's called personalized medicine because we have to understand your individual genome, your individual function, and then use the version of the drug that is addressing that particular combination that you have. Yes. And of course, every human genome is personal — otherwise we would look exactly the same. Even twins — even identical twins — have some differences between them.

Yes. But yes — personalized is called that because we target it to your person. But that doesn't mean that combination won't work with anybody else who has the same combinations of genes that you do.

0:39:10 Justin McMillen

I see. What's the advantage of this? So this is part of why you and I are even sitting here.

0:39:10 Vahan Simonyan

Yes. So personalized medicine — is there an accepted definition of that? There is. The definition — I mean, obviously everybody can Google it, but the definition is that decisions on a treatment — I'm going to use human language, there's a formal definition — decisions on treatment depend on the personal individual characteristics of the patient.

0:39:36 Justin McMillen

Okay. So is that just — we're going to gather as much information about you as possible, preferably genetic information, and with that, understand how to best provide you with some form of health care?

0:39:36 Vahan Simonyan

Perfect, perfect. But that's a huge challenge — a humongous challenge — for multiple reasons. Number one, that's a lot of data. But now we have the technology for it. Number two is not from the area of technology — it's from the area of clinical data.

Governance, contractual obligations, patient privacy — all of the things that are prohibiting scientific research. Because even though data is generally generated, even though data is a treasury of information, moving it and using it is very, very difficult for completely non-scientific and non-technological reasons. And these reasons are completely related to privacy and an exaggerated feeling of security, and other contractual reasons — data governance and data oligarchies.

0:40:32 Justin McMillen

There's a different reason we can discuss if you want. But yeah, I mean, I think you know how I feel about privacy. Obviously we're all big fans of privacy. I think there's a market for data — it's an important piece of that. There's a lot of people making money on it.

And I think that's a huge piece of it. But if all this red tape was removed, would we see this massive change in health care and people's overall health? What would happen to innovation if we could just take and analyze data and look at it? What would happen?

0:41:16 Vahan Simonyan

I think if you remove the red tape — and by red tape, I mean if real scientists could get easy access to real data — the explosion of science would be far, far bigger than it was because of the internet. Something would happen in health care that is multiple times comparable to the appearance of the internet in human life. Every aspect of human life has been touched by the internet. Everything exploded. I think if we let people use data...

And I'll give you another example — NOAA, the National Oceanic and Atmospheric Association, was generating all the weather data all of the time. And that data was sitting somewhere on the servers. And then I don't remember the name of the scientist who decided, I'm going to put this data into public consumption — and they put it out, available to any academic researcher who wants it.

And students from different universities started using that data to test their physical models of weather prediction. Voilà — in two years, our ability to predict weather increased hundreds-fold, just by the mere fact of letting young students play with the data. And then in two years, we had a complete innovation on how weather prediction works. And now even when they say it's going to rain — there's a very big chance it's going to rain.

So in two years — I think the biggest thing we will see if we cut the red tape — we will see two things. Number one, it will have a huge impact on the cost of drugs manufactured. The cost of diagnostic things manufactured. Second, we will see an explosion in academia.

We will see a lot of different interesting algorithms and correlations found, and it will be heavily mined. And I'm coming to you — I truly believe we are decades delayed because of the red tape.

0:43:25 Justin McMillen

Thank you.

So amazing. That's — I mean, what's sad about it — and I, you know this, but there are people right now out there — maybe people listening to this — who are sick and there are answers to what's going on with them.

0:43:50 Vahan Simonyan

Yes.

0:43:50 Justin McMillen

That exist. And because of regulation — and again, I want to be clear, both you and I are huge fans of privacy. So that's not what we're talking about here so much. It's really just — we have regulated ourselves into a nonsensical mess in a lot of this stuff, and it's fueled probably a lot by data brokers and different things.

0:44:17 Vahan Simonyan

And I'm not making anyone bad — everyone's doing their thing. But we've put ourselves in a situation where — you said we're decades behind.

0:44:44 Justin McMillen

Yeah.

0:44:44 Vahan Simonyan

And people are dying. People are dying because of that. In fact, just to reinforce what you just said — we are for privacy and regulation because we designed the technology that is ensuring even more privacy while allowing things to be used.

So I — we have discussed this before, meeting on a couple of occasions. We were funded — our entity was funded — on an ARPA proposal. ARPA is like Advanced Research Projects Agency. There is, for example, DARPA — Defense ARPA. DARPA was the reason why we have the internet, why we have cameras, why we have satellites — some of the most amazing discoveries.

0:45:10 Justin McMillen

It's also the root of all kinds of conspiracy theories.

0:45:10 Vahan Simonyan

DARPA. Well, yeah, but I bet...

0:45:10 Justin McMillen

Yeah, we can definitely confirm the good stories.

0:45:10 Vahan Simonyan

Oh I know, and that wasn't the right area, but ARPA-H is the same organization — a similar organization — for health care. ARPA for health care. So they have funded, about $10 million, to create this new technology.

It's called FEAST — Federated Ecosystem for Analysis and Standardization Technologies. Let me explain what that means. So you know, like every time you go to a doctor, you leave a trail of data in the clinical system where you attend — it's called an electronic medical record. EMR. You're typing in data, data being typed on a computer. There are EMRs — electronic medical records — that are accumulated.

Every time anything happens to you — you go to a pharmacy, you go to any interaction in the clinical space — you leave a trail. All of this data is a huge value for health care and for wellness. Yes, we know that. But nobody wants to leak the data because you are concerned about your privacy. And the hospital is afraid to release and share the data because of similar privacy issues.

God forbid your data leaks and somebody finds out you have this disease. Well, for you and for me it might not have any implication, but some people might be really upset by it. And so on and so forth. Privacy is a really important subject. So data never leaves the hospital. On one side you have the patient saying I don't want everything private — I have no idea what it'll be used for.

0:46:43 Justin McMillen

And then on the other side you have the hospital being afraid — or the doctor's office being afraid — that they're going to get in trouble. Yep. Either from the government or get sued or something. Protection on both ends.

0:46:43 Vahan Simonyan

Protection on both ends. So — but then now, let's imagine I have a pharma company who wants to design a new drug. And I need data to even start understanding how this disease works. Let's assume your data as a patient would be really useful. If I go to the hospital and say I want to use that data — there are two scenarios.

Yes. Number one — I'm shown to the door. Get out of here. Okay. Number two — I can apply to do that research based on that data. But then there is a story to that. In 1985 — I think — the Health Insurance Portability and Accountability Act. Yes. HIPAA. And so it tries to protect the human health care information. But you have to remember — that was when it was paper. Everything was paper.

You literally had to make sure that your paper records are not accessed by people who are not authorized. Yes. And once a person has access to that paper, there is no way you can secure it. Right. Yes. But now we live in a completely different century. Literally, we have the means to protect information, share and protect it.

So the other alternative — the hospital might say to you — you have to do this in the Institutional Review Board. Let's make a bunch of scientists go and look at what you are going to do with that data. You have to write a proposal, and scientists will get together, analyze what you intend to do with the data, and give you a yes or no signal.

Can you have the data? Which is not a big problem by itself. Yes. The challenge is — the moment they release the data out of the hospital, now you have to remove all of the patient identifiers. And let me explain what that means. I know you know this, but let me just reiterate from the audience's perspective. Yes. So let's say you have a blood pressure.

There is a number — the blood pressure number. Who does it belong to? Yes, it belongs to Justin. Let's say your identity number is 1-2-3. Yes. And then you have genetics done. How do you know it's Justin? Because it's associated to 1-2-3. Let's say you have some imaging done — X-rays or CT scans. How do you know it's Justin's? Because it's associated with 1-2-3.

And so 1-2-3 is replacing my name. So my name doesn't exist anywhere. Yes.

0:49:09 Justin McMillen

I want you to tell me how 1-2-3 could somehow lead back to people knowing it's me. You keep going with your point, I'm just — I'm sorry. Yeah, I'll come back to that point, remind me of that. I'll come back to that point here.

0:49:10 Vahan Simonyan

So — so — it requires that if data is to be released, we have to de-identify the data so it can never be traced back to Justin the individual. Yes. The unfortunate reality is the anonymization procedure — we call it anonymization because we're removing the names — but you're also removing anything that connects data together. So we end up releasing data that are disconnected.

I might get an image — it's the 13th image in a set. That might belong to you. But then I get the genetics and now you're number 27 in the other data set. How do I know it's the same person? Yes. I mean, it's difficult. There are some ways to try to keep a certain level of connectedness.

That situation is much worse because — let's say you go to a different hospital. And Americans move all the time. Yes. And now you're a completely different patient. Your data is not maintained consistently in one place to analyze and make sense of it.

0:50:32 Justin McMillen

It just clicked for me. So it's not just that — like you go to Johns Hopkins and you pull a data set from a guy named Bill, and okay, we've got his radiology, we've got his blood — we've got all these things. We're going to go ahead and rename Bill "115" in one data set. That's fine because you can still correlate, yeah, it's the same number. But once Bill is in a different EMR — that's it. Because you have to remove Bill's name. There is no way to make these connections.

0:50:51 Vahan Simonyan

Yeah, so... okay. Keep my hospital where I go. Well, I'm — for more than 40 years. Obviously everybody as a patient has had seven EMRs in a single hospital system. So when I move between floors in a four-floor building — seven EMRs.

0:51:16 Justin McMillen

Here? That's crazy.

0:51:16 Vahan Simonyan

So the data is difficult to connect even within one hospital system. That's the challenge. Yes. It's a big challenge. Now, just to understand who you are from a health care and wellness perspective — I also need to throw into that your activity. What type of physical exercises, what type of diet, what type of socioeconomic conditions are affecting you? What type of psychological factors are affecting you? You being you and your health depends on so many variables. And now they're all in different places.

I saw you earlier this morning. Most probably that data is in an app server somewhere else with no association whatsoever with other clinical factors about you. Yes. Which means that we're losing an opportunity for that information to be valuable in terms of making good decisions about health — the health of the country. And that's called segmentation and fragmentation. It's segmented by types of data, and it's fragmented by location and governance.

Who governs, who controls the access to the data. So I was talking about the ARPA funding. So we got funded for a specific project to generate a new paradigm, which we proposed. It's called FEAST — Federated Analytics paradigm. For short, FEAST. Federated Ecosystem for Analysis and Standardization Technologies. So it's very simple. You can explain it like we're doing around here just now. Yes. And we are lazy — scientists are usually very lazy. We don't want to do things multiple times. And if it's really difficult, we just don't do it. Imagine trying to get access to this data, that data, that data, that data. And these regulations are really tough — even though important, they're tough. I cannot move data to a centralized location to analyze together.

Well, we cannot move the data to computation — let's move computation to data. That was the proposal in front of ARPA-H. We decided with this technology, with virtualization paradigms, we can take a computer algorithm running and move it from one computer to another, to another. What if these computers are in different clinical systems? Instead of taking the data to my computing server, I can take the algorithm to the data, use the data — whenever it needs another piece of data in a different hospital, it just jumps there, uses that data, jumps there, uses that data.

And in a prior discussion between us, you gave this wonderful example. If you allow me to repeat that example — imagine there are these libraries, but you are not allowed to take books away, but you are allowed to read inside of the library walls.

0:53:52 Justin McMillen

Yes.

0:53:52 Vahan Simonyan

And this is a wonderful example, because let's say I want to learn about a subject. And four different libraries have important books. I need to read them and I can't take them home. I will take my sandwich, my juice, go to the first library, read the book, put the book back into the library, go to the other library.

All I carry with me is the mental state of my thinking process. I've learned it. Mental state. Yeah. Go to the other library. Do the same thing. Go to the other. Do the same thing. If I need to return to another library for another book, that's okay too. So this way we did not take a single book from any library, but we learned everything there was to learn.

So the same way — the analytical computation. Let's say you have an AI algorithm. Yeah. The algorithm needs to get access to this data, that data. It finds out — we have a way of telling it — that this data is in this hospital. It goes there, does its analysis, gets a half-digested data set, and then moves to another, moves to another — different patients' data sets — and comes back with an analytical result.

0:55:01 Justin McMillen

So you're not — okay. Let me see if I get this. And we've talked about this a bunch, so I think I got it. I think I can explain.

So going through the front door — the way everyone's been trying to do it — that's highly regulated. Give me the data. Only one data set here, got to do it multiple places for one person. But give me the data — before you give it to me I have to... strip the names off of everybody.

Which means that in large data sets I can't connect things together to correlate. So — you're sitting back thinking, how do I solve this? You get funded from ARPA-H — DARPA's brother, or sister, whatever — which is set up to do moonshot-type things, solve major issues.

Major. Right. So the government is saying — we have to have a solution for this because people are dying, there's an issue. We can't use this data because of all the laws that exist that need to exist. We need someone to come along and produce a solution for us that allows us to get to the end result we're looking for without violating any of these things.

So you — being a physicist, a first-principles thinker — you're like, what's the end state? The end state is understanding and knowing. Right. So you're like, as long as we get the answers, I don't care if I have the data sitting in my server. I don't care if I have to — you know — I don't care. It's about the end-state answer.

So how can I do it? So you said — wait a minute, I'm going to analyze it. I don't need to do it at my house, in my office, in my data center. I can build something that can go to the data, gather pieces of information about that data, and bounce from different data sets — read the books, get the full picture.

0:56:47 Vahan Simonyan

Yep.

0:56:47 Justin McMillen

And then come back with the analysis already done.

0:56:47 Vahan Simonyan

Wonderful. That's what it is. So it is built. It is operational. We have demonstrated it for a few use cases. Now we are running the cancer use case — we are moving between seven clinical organizations, massive organizations with tens of thousands to hundreds of thousands of patients each — and analyzing different types of drug resistance in cancer. To do this type of analysis you need large populations of patient data, which are spread across the country, East Coast, West Coast.

We are in Kaiser Permanente, we are in Weill Cornell, we are in Vanderbilt, we are in Georgetown. And all of this — the computation jumps from place to place, gathers and analyzes data, comes back and tells us — these types of patients will be resistant to daratumumab treatment, but the other types of patients will be okay to give daratumumab. Just to give you an understanding of the level of impact — 70% of the cases when people are given cancer drugs, the drugs are not helping them.

70% of the cases — depends on the drug, but on average, cancer drugs 70% of the time are not helping patients. They are tens of thousands to $100,000 for treatment. Yes. And now 70% of the time it's useless. But we have the data to understand — hey, you, patient, I'm sorry, but this drug is not going to help you. It's going to hurt you without helping you.

There is a risk without benefits. Yes. But to do that, I need to know the patient's data and compare that to other patients' data. Now imagine this algorithm by itself may have a huge economic impact — just in avoiding giving people drugs that are not useful to them. In fact, you are pushing them further down into suffering without the benefit.

Because if people go to chemotherapy and have side effects, then you have to take care of those side effects also. This is like a machine that is not going well. Yeah. Yeah. So this type of algorithm that is allowed to do this — it's a huge economic impact. So that's the cancer use case. We are also working towards the medical device use case — devices that are used for aneurysms.

And we also are running the Lyme disease network right now. We're trying to understand chronic Lyme, versus Lyme, versus other diseases which have similar symptomatic cases — jumping from hospital to hospital, trying to understand what are the preconditions that cause somebody to get chronic Lyme — it's like long COVID, except much worse — versus somebody who gets well from Lyme. And so on.

Again, I need thousands and thousands of patients from different locations. We are using this technology today. FEAST is live. We have a Lyme network. We have a cancer network. We are going to do — and you know, we can discuss this if you want — the addiction research. This technology exists today. And then we are actually trying to use it for another use case that I cannot go deep into details about for obvious reasons.

Biological surveillance is a very big issue. Because today we can detect a pathogen here, pathogen there, pathogen there. Let's say a clinic gets 20 patients with the same disease. That's interesting. If it's two interrelated, then CDC or a state lab detects it. And then maybe Port Authority has brought that in with some products — some products have been shipped and Port Authority hasn't detected it. Or maybe it was detected.

There is a discontinuity of data. And the reason is because they don't share data — the Department of Defense doesn't share data with CDC, who doesn't share data with clinics, who doesn't share data with diagnostic labs, etc. The same challenge — we are smart locally, we are stupid globally.

1:01:03 Justin McMillen

Yeah, no, it makes complete sense. It's like — and this is a perfect thing that things are emerging this way too, because again, we're getting more advanced, like you said. When you started with NIH and FDA, they couldn't even figure out how to store...

1:01:03 Vahan Simonyan

Oh yeah. Yeah.

1:01:03 Justin McMillen

So now look at how things have advanced. It's already to the point where we can store it. And now we have — you know, the biggest complaint with health care professionals is — I'm spending more time charting and documenting. A lot of that's driven for a lot of reasons — insurance. And there are challenges there too. Because perhaps things are not recorded in a way that's quality.

Right. So your point — I have a question going back to earlier. Talking about if you send out this little FEAST to go read the different library books across different data sets, across different hospital systems — can you make it about one single patient?

1:01:44 Vahan Simonyan

If the patient has data in multiple — yes. You can. I mean, well, you would know — if your intention was to do that, then you know the identity. So you're fine anyway. Data never moved here, but that doesn't prohibit that. I actually — sorry, I'm not — let me say it a different way.

In order for you to have an understanding of the different connections between different data sets, you need an identifier. We're not using a number according to that — right, we're still going with the name. So there's a ground truth that makes this data connect to this data, connect to this data from — you know — this record is also the same person.

The individual subject is the same in these two places, and the subject is the same here. How can FEAST go around and know the subject is the same? Is it just pulling the name? And then also — how come it can't come back and tell you, hey, by the way, I just saw Bill's stuff in five different environments, and did you know he has some disease?

1:02:48 Justin McMillen

So you're touching two points at the same time? Yes. And one of them — I love specifically — is the governance of the data and who owns the data. And who allows me to see it. If the patient is the owner of the data — which in this country, it is. And the patient has a right to access their data in most of the states.

1:03:15 Vahan Simonyan

But the patient doesn't govern the future of the data or the appearance of the data. Really, if the patient — you the patient — owned the data, it would be your choice to give the algorithm access to certain elements of your data, which may be all over the country. Yes, they're all over the place, wherever they are. The unfortunate reality is today's rules...

Although they require you to be able to request and get the data — they don't enforce the means and don't impose timelines or the quality of the data exported to you individually. Let's say I'll give you an example. Let's say you did a CT scan somewhere and then you go to that radiology lab. You say, I need my CT scan.

They'll give you a CT scan. Perfect. It might actually be on an old slide. It might be on paper. It might be on a CD. And go figure where you can find an old CD reader just to read that. Or it might be on a USB. And that doesn't mean much for most of the patients — because most of the patients who go to a radiology lab are 60 and above, people who are not very actively involved in technology and software. Even looking at their data is difficult because these are very highly specialized formats.

So you have the right but you don't have the means. That's very important. You have the right to get access to your data, but there is no regulation giving you the means to understand the data or to actually truly access the data — not just having a USB stick saying, hey, I have my data, which is useless and sitting on the counter somewhere.

Yes. So regulations require you to be provided your data, but don't say how long it will take. And the hospital will always say — we are very busy. So you don't know when it's coming. And the hospital can also provide it in formats that have no standards. Yes. All right — ten different ways I can present that information.

I can print it in a PDF or I can give you a PDF file or something. There is no standard on the means to deliver data to you. This is a problem because if you don't truly own the data, it's in their hands to decide who to share it with. For you, they'll give it in a way that's not really convenient. It's going to be in a language you can't understand.

Yes, but — if now someone is acting on your behalf and saying, hey guys, listen — Justin gave me informed consent to use his data, these particular data elements, for this particular study, during this period of performance. And here is the signed information from Justin.

Then it can go in different scenarios — they can say, oh, now you have to submit a contract with IRB. Yes. You have to go and prove that you are not going to make any mistakes, and things like that. Which is not a big problem — unless you recognize there are thousands and thousands of IRBs across the United States. So if you are an organization trying to get data about Justin — sort of, across houses — you need to do a thousand contracts with different organizations.

You're asking why drugs are so expensive — this is one of the reasons, because data is so expensive to get through these contracts. But if you own it — if you change the system — you don't just have a right to access, but you own it. And then hospitals must release the data. And then if Congress and HHS — they have an OCR, Office of Civil Rights — if they enforce the standards and they demand standardized and time-constrained delivery of that information to a third party on your behalf, you solve the problem.

You solve the problem. Also — and this is the second part of the question I started talking about. If Justin is in his hospital, in that hospital — you provide consent. You tell the name of the hospital. I already have the protocols that can retrieve any information from anywhere. But what I don't have is your consent in my system right now — and the hospital's willingness to play along. The hospital will be stopped.

They have no rights to stop it, but they will make every effort not to be actively doing it, because it's expensive for them also. Sure, it's got to be. Yeah. So okay — so this is what FEAST addresses. We talked to legal teams in the HHS domain — people who are working with Health and Human Services, pretty much health care lawyers.

We understand the data privacy and things. So this is doable. I think it is interesting. Yes. All my life I did science that required me to use not just technology, but now we've done the technology. Now it's required me to understand contracts and legal legislation. So it's only one more step left. One more step — change the legislation.

Because before, we didn't have the science — we didn't need that much data. Now we have the science. We need the data. We didn't have the technology. Now we developed the technology. Now we have the data. But to get that technology to work, I need access. Access — I need contracts, I need the legislation to change.

1:08:56 Justin McMillen

I want to be clear for people who are listening so they understand what you're fighting for — what it's not... what you're fighting for is not — it's for individuals in this country to be the true custodians, the stewards of their own — yeah — to have access to the custody chain of their own data, so that they can give permission to people like you or others who want to use that information in order to improve the health of our country. And you can do that with the full consent of the patient.

1:09:23 Vahan Simonyan

Exactly. And in theory — I mean, I think people in this country already think that they're the ones that get to make that decision. They don't realize that they don't actually own it. They own the right to it — but it's all constricted.

1:09:47 Justin McMillen

Yeah, it's all constricted. And it's very — because there are forces that you don't know about — they don't want you to have the ownership of the data.

1:09:47 Vahan Simonyan

And there's no conspiracy theory here, of course. Yes. I mean, today there are huge transactions of data between — I like to call them data oligarchs. These are clinical research organizations, or EMR organizations — providers like Epic and others — who benefit from sitting on this data. And I by no means want to place blame on them.

There's a business. They perform a very important function in the entire economy. But if patients owned the data, these organizations might lose big — because today, for data transactions, regardless of your willingness, you have to deal with them. And it is an expensive, very expensive endeavor. For a single clinical trial in Ireland today — like a gene editing clinical trial — it's $6 billion total before it hits the market. About $1.2 billion of that is the cost of data.

For a single drug — $1.2 billion is the cost to generate or access the data.

1:11:10 Justin McMillen

Okay. So now I'm going to ask you a question. And this is — I know for the audience, you and I talk a lot and this stuff is — I get this. And I remember the first time I ever heard you talk about this stuff, I sort of was like, there's a lot there.

1:11:10 Vahan Simonyan

My head was swimming and it took me a while to organize myself around it.

1:11:15 Justin McMillen

The end result where we're going to get to in this conversation is, I think, really important. And it's simple — around why this matters so much for the people of this country and the future of our health. So hopefully, if you're still listening, you're getting through all of this. Not that any of it — it's all interesting, but I know if people don't understand it, it could cause them to be sort of tuned out.

But if you keep listening, I think you'll get to where this is going and why it matters to all of us so much. And here's the first point. You just said it costs over $1.2 billion — that's just for the data. Okay, so what he's saying — if there's somebody making $1.2 billion...

Okay. So I'm curious for anyone who's listening — when was the last time you were paid a penny because of this data that you're supposed to be the owner of or have access to? And I think the answer to that is probably never. So who's making all this money on selling data? Should we talk about that or should we not?

1:12:20 Vahan Simonyan

I mean, let's stay out of names. Stay out.

1:12:20 Justin McMillen

So somebody is buying and selling information about me?

1:12:41 Vahan Simonyan

Yes.

1:12:41 Justin McMillen

Across the country?

1:12:41 Vahan Simonyan

Yes.

1:12:41 Justin McMillen

And if I wanted it, I couldn't even get it. Even though I'm supposed to legally have it. Due to privacy that's making it difficult. That's... the machine is broken.

1:13:01 Vahan Simonyan

Yeah, it is.

1:13:01 Justin McMillen

And I don't think it's nefarious stuff, actually. It's not nefarious. Yeah. I don't think some evil person is doing it. It's just... it's just not the natural way of things. Right? Like capitalism — somebody got smart and was like, oh, here's how we can do this. And you know, but I think what's interesting about this is — you said it costs so much money and we have these problems.

And if we solve them, we get healthier, and potentially change the entire — if we can reduce this data cost for people developing things, drugs are going to be developed for more rare diseases that otherwise wouldn't be developed. Yep. They'll be cheaper for everyone.

1:13:48 Vahan Simonyan

Okay, just to make sure this is not just sounding like a utopian... yes.

Please. The structure of why it's so expensive to do a proper study and to bring a drug to the market — number one thing you do, you go for preclinical data. So you start accumulating thousands and thousands of data from patients having the disease that you're trying to cure or treat.

That involves working with hospital systems, finding out which hospitals have how many patients and things — going through contracts, going through all of these things. We already discussed — an average cost of just an EMR, your medical history, can range per person from $50,000 and more per patient. And you need thousands and thousands of patients.

1:14:41 Justin McMillen

So somebody's paying $50,000 for a medical record, for important — so-called "hot disease" areas. Yes. Yes. Of course. Is there any — if you don't want to answer this, it's okay — is there any way to have that $50,000 go to the person who actually...

1:14:41 Vahan Simonyan

Oh yes, of course. I mean, that's the idea. Once we have patient ownership fixed, and it will be the only way to access the data — and I'll give you the... I mean, these ideas, health data markets are — I know you have stuff around this, I don't know where you got to — tell me. We it's okay in this.

1:14:55 Justin McMillen

You can say just that. That's a boundary, don't go there because I — no no no no, we can't, we can't go up. So if you don't name the companies — okay.

1:15:13 Vahan Simonyan

But I also don't want to talk about some things you're innovating on that are incredible — that are going to change everyone's lives. That's how it is. If I move in that direction just say shut — I've just — but that's okay. I mean, we can talk about this. Yes.

And I'll give you an example. Like, let's say if I buy EMR data — and some of our clients we help to purchase that and give us to analyze. And so let's say like $50,000 for the patient with a long history that's normal to get, plus some genetic information, plus some imaging information — there are all these cost factors. In a clinical trial setting, it takes me $8,000 to $20,000 to do genetics.

If that's already done — if I come to the patient and say, I'm going to give you $2,000, let me use it for this research — the patient says yes. Me as a company, I got it for $2,000 instead of $8,000 or $20,000, depending on the type of experiment. And then I can do it multiple times. Because you know, a direct monetary beneficiary of that transaction between me the researcher and you the individual patient — you are incentivized and I am incentivized to continue this type of relationship.

Come again. Come again. Come again. And data is a non-zero-sum game. It's not like when you give me the right to use that data, you lose the data. You don't. You still keep the ownership. In fact, if I'm a company, I don't want to have ownership, because if I have ownership of the data I can use it once — and then next time I still need consent from you. But because you didn't make a dollar off it, you're not incentivized to give me the consent.

Now I have to go recruit another new patient, which is much more expensive. So in a way this is benefiting the patients. The hospital systems actually — if you build the value chain — every time I spend $1,000 for this data piece, let's say the patient gets $500, the doctor gets $200, the hospital gets $300.

Everybody benefits in this schema. Everybody benefits. Because otherwise, if the person who can also write the research is not the patient who owns it — now we have two individual things governing my ability to access the data. And then if this individual is not incentivized — that's it. My machine is broken. There are no incentives for people because there are no monetary gains or some other gains.

You know, we have done this experiment in different countries — actually paying people back for their data. In the US it's much more difficult because of all the complicated regulations. So imagine an old grandmother getting a call one morning saying, because we used your data, now we need you to tell us where to deposit the $8,000.

Her monthly pension was $100. Now her data was valued at $8,000. And you know what she said? She said, can you use that money to help kids? She didn't even want the money. I mean, so... but of course, if people want the money, it will be transferred. That's another type of incentive. You can use the same money in some states — in some governments, you cannot pay people for their medical data because you don't want to incentivize them to do extra medical procedures for generating more data.

But then you can use that money to pay against their medical insurance or medical treatment co-pays or something. So there are all types of ways you can benefit. This is why — and this is where you and I really started going crazy, because when we talked about this — the first time I heard about HIVE, I was so fascinated because, you know what I do, and my big challenge in our work is we are treating human beings multi-dimensionally, and we're trying to pull information in from many different sources about biology, psychology, and the social dimension of addiction.

1:19:15 Justin McMillen

And the idea is — if we could pull that information together, analyze it, find correlations — we could be way better at what we do. So you're telling me about HIVE and I'm thinking, this is incredible. And at the same time I'm thinking about self-directed health care, which is what I see as a future of health care in general. I think the future is that we're all going to be able to use AI-type technologies that we can feed information about ourselves into — and live in a place like HIVE that's sovereign and protected, non-governmental — and then continue to improve upon that by feeding more information, so we can get meaningful information about how to direct our own health.

So I'm thinking about this — and I only know pieces theoretically — and then you start talking about this. And I realized very rapidly that the key to this is everyone has to own their own information. The only way the public is going to want to own their own information is if they understand that it's valuable. There's an entire market for this.

1:20:08 Vahan Simonyan

So if people didn't catch it — I'm sure people have — but the amount of money that's being spent on data, the amount of answers we need to find based on that data, are so vast and so great and so transformational that there's a market and an economy around it. There are issues, and this could be something that's solved socially on such a broad scale...

1:20:28 Justin McMillen

...that everyone is able to make money off of their own health care information, with their own sovereign choice to do so. And in the end — and for every study, it's only one time. Which means — and you told me this — it's almost like a bank account that keeps rolling. You spend money, but it never goes away.

1:20:49 Vahan Simonyan

Yeah. And so — why that's interesting is... the sicker people — by the way — the people who are sicker and have the most problems would have the most valuable health record. Which means that you create health care equity in a very healthy quality. And I'm sorry — you offset the kind of damage that's happening financially to people by giving them more resources.

And also, all of that machinery will have the potential to decrease the cost of the drug from preclinical stage all the way through to the market. Think about this — a drug is now approved by FDA. You are the company. What do you do? You start running advertisements all over the place saying, hey, if you have this condition, come try my drug.

It's okay if it's a broad drug — most everybody has the condition. But if it's a targeted drug — why would you go and start running a TV advertisement all over the United States, which is very expensive? If you can just do a search, find out exactly where the patients are who have that particular condition and cut to them.

Do you know how much that saves — another billion dollars from the advertisement cost alone. Because if I can go to a destination system — and we're trying to build something similar, yes — you go and search. Okay, I have this drug which addresses this condition. I need to find all of the doctors who have patients who might benefit.

That's it. You found them. Instead of going regional campaigns all over the place, paying hundreds and hundreds of millions of dollars in blunt advertisement, taking doctors to Hawaii — doing all of these things that's how companies work today. Sure. That's a huge chunk of money. You don't recognize it now — we're kind of out of medicine, out of science — but it's a huge chunk of money.

All you need to do — find me that patient. That's it. And you also save — yes. Saving before you start the study. Saving once you're finished. And saving while you're doing the science. I mean, and ultimately saving the consumer. I think about — you're a biotech, you have investors, and you're spending so much and the only way the margins go up is by the number of drugs approved in the market. The percentage doesn't go up. It's all very difficult.

Imagine if you're showing your investors that your costs are coming down significantly. Yes, I mean that's a huge factor. Now that means more investment into it — because the resources that are not necessary to spend on these unnecessary things can be redirected to other areas of research. Imagine the exponential value of this — a new way to do research.

It's amazing. It's a crime not to put our lives into trying to make this happen.

1:24:02 Justin McMillen

Yeah, no. I think I mean — that's — I think Lukian, our producer sitting over there, is probably thinking about — I wonder if I could get $50,000 for the last time I went to the doctor. Right. Imagine that. Imagine if you could take your medical record, go hit up one of the places that you've been — whatever hospital or something — and just say, I want my medical record, and then sell it.

1:24:02 Vahan Simonyan

Yeah I mean, that's happening. I think that's what people don't know — this is what's happening all of the time already. And the public — in some of the cases now, like 23andMe was caught selling data without consent and things like this. It's a subject. And it's really important because the questions are big right now because AI is moving fast.

And that's a broad term that means so many things. We talked about this earlier. But supercomputation, the ability to manage large amounts of data, algorithms — all the things that are leading us into our future are changing the game. And the question of who owns information is really important. We've already seen this happen with social media, right?

1:24:44 Justin McMillen

People tracking our behavior without really getting our permission. Everything's free — why is it free? Because they're selling the damn information. And nobody that's contributing is benefiting. In fact, we're becoming dopamine monsters, right? We're just scrolling through all this stuff — and it's considered addiction, you know? Yeah. So it happened once with our social behavior information, and now it could happen with health care information.

1:25:07 Vahan Simonyan

And the government may — yeah. I think it's happened in the UK. I believe the government has taken control of health care data in order to solve these patterns. They gave up. They said there's no — I'm not saying they gave up — I don't know enough about it. But the point is that if we don't allow this to happen in the right way, we could move to a dystopian situation where the government knows all of our health care information.

They have access to it, it's not private, there's no benefit to the public, and perhaps the government is making money selling it to the same people who are continuing to develop. And perhaps drugs stay expensive. That's a pretty dystopian, terrible way to look at it. But I mean, in America, we have a history of not trusting big brother government.

1:25:58 Justin McMillen

And that's the last thing we would want government to know everything about — including our medical histories. Yeah. Especially when they're making decisions about whether to insure us or how to... I mean, so many things are changing. Yeah. You know, I'm — and I'm very much so in this — because of the fact that we treat addiction through all these dimensions, I have this belief that the future of all health care is going to be looking at treating the whole person rather than in silos.

I mean, science is beautiful, but the constraint is — we have to isolate variables in order to do it right. So all these therapeutics are really focused on single variables living in a vacuum that don't apply to the actual complicated orchestra of parts that is a human being. So we treat people multi-dimensionally. Okay. So why does that matter?

Because I think the computation now has the ability to actually look at all that information. And again, it all comes back to data. So how can we gather enough information to know how to effectively treat somebody? Where does that information live, and how do we use that information to make decisions? So I'm obsessed with this idea that — because now everybody's going on ChatGPT, putting their bloodwork in — maybe dangerous, maybe not.

1:27:08 Vahan Simonyan

I'd probably not advise that. But people are finding answers to common problems they have using LLMs. You know — that's problematic. We talked about this. But like, you also said — there is no doctor on planet Earth right now that could keep as much information in their head as an LLM. Okay. So let's consider this — where is this going?

1:27:32 Justin McMillen

I don't think people want to look at it because we're all deeply embedded in our own careers and our own lives. But as we're all sitting here going, I'm just going to keep doing my work — we have these things that have more information than humans do. And the public — I don't know. This can be top-down.

The public is demanding to use these things. They're doing it on their own. Yeah. So as they're doing it on their own, they're gathering information. So I've been saying this for a while — and as we first met, this is what I believe to be self-directed health care. This is people using these tools to start making decisions on behalf of where they want to take their life in terms of health.

So going back to this again — without data and storage and control and access, it's really hard to do that. A future where I have access to all of my data allows me personally to be sovereign and make decisions about my health in a way that could never have been done before. And so that is what I believe to be the future of medicine.

1:28:31 Vahan Simonyan

And I think you're talking about — I concur. Yeah, yeah. It's going to happen. The world is going to accept this, or it's going to happen without us. But if we don't do something, we have to make some decisions around it soon.

1:28:48 Justin McMillen

So then — what's really cool when people don't know how you and I know each other. So after these conversations happened, we started talking about the work we do. So I had to explain to you to understand why — and you started to say you were looking at numbers around addiction, and we were talking about how we could work on a project together. And this is where we'll go into CARES-IT. So it's appropriate now.

1:29:06 Vahan Simonyan

Yes. It's interesting. Okay. So we're talking about addiction and our country trying to treat addiction and what's happening in addiction. And you know, I'm — you learn faster than anybody I've ever met. But I'm just downloading all these things to you, and I think my perception is you're sort of baffled by just how — like you talk about the hammer in a...

1:29:28 Justin McMillen

Yeah.

1:29:28 Vahan Simonyan

Crystal shop. Yeah. It's like — there are no agreed-upon targets. Targets being the end state we're looking for to determine whether someone is healthy. We can't — we say sobriety instead of disease remission, like you do with all other chronic diseases. And I think you quickly realized that — and things I know and others in your field know — this particular chronic disease is costing the country more money than any other.

1:29:56 Justin McMillen

Well, be careful with this — the total cost burden on the country is greater than any other chronic disease, because of all the loss of life, loss of productivity, incarceration — all these things. We have this massive issue and nobody has really taken the time to understand it.

Yeah. And I don't want to take credit away from people — there are amazing researchers. I mean, we have at NIDA — the woman who ran NIDA, Nora Volkow, has done some amazing work. There's some incredible stuff out there. But broadly, we haven't given it the same kind of — we haven't put the same resources towards understanding the nature of addiction, how to effectively treat it, as we've done with all other chronic diseases.

It could be because of stigma. Could be because we didn't have the computation to do the kind of analysis to look at all the factors that contribute to somebody choosing to destroy themselves with substances versus not. And so that was stuff we started talking about.

1:31:01 Vahan Simonyan

Yeah.

1:31:01 Justin McMillen

And then you brought up ARPA-H. And can you share a little bit about what you remember about the early days of CARES-IT? And we're not early because it's only been a year, but yes. But maybe — before I do that, you talked about AI and analyzing the data and people trusting AI and things. And so I want to make one more point before we move forward, because we are at a point of a junction that is very important for decision-making for human health.

1:31:26 Vahan Simonyan

So let me give you this point. So most people, when they know how amazing AI is, think like — AI will solve all of the problems. Yes, we hear this. AI does this, AI does that — every time you touch the thing you recognize it has a huge potential. The challenge is the following — AI, all of us...

It tries to analyze the data that previously exists. It finds patterns in the data or in texts, and then reproduces the patterns. AIs are not really smart. Not yet. AI is stupid. I can say this — hopefully one AI is not going to cut me off of this, right?

1:32:14 Justin McMillen

Yeah.

1:32:14 Vahan Simonyan

But what it is — and what AI can be — is very smart. But to be smart... think about what is a disease. Disease is a correlation between thousands of variables describing a human. If you are describing a very complex situation, you need a lot of data. AI without a lot of data access — it's not a good AI. It's an overtrained beast. A lot of confidence, not much intelligence.

Yes. Imagine if I have ten people who only read the Quran or Bible and nothing else in the world. They may be the ten smartest people in the world. They will be very trained, but they'll come out and the only things they can recite and be good at is what they have read in that book. The same way when AI is not provided a complex, very complex set of data on a broad population spectrum — which is human diversity, genetic diversity of humans — AI is going to be just like the Bible reader or the Quran reader who just keeps reciting what he has already heard.

So the difference between AI that is intelligent and — I mean, it's a conundrum. Limbs and things... so it's motility. It's computer motility. The FEAST AI makes a decision about what other data it needs to read. Data is there, because of regulation I can't move ahead. Therefore I can only train on small data sets that it has access to.

Imagine if now AI becomes more mobile. The difference between mobility and motility — mobility is something I can move, it's movable. Motility — if that thing makes its own decision where to go. Yes. So when AI is combined with motility — computer motility — that's what FEAST technology provides. Now we call it motile intelligence instead of artificial intelligence. The difference is humongous. Because now that AI has potential access to all data — not just this bunch of data or that bunch of data. And we all hear about hallucinations and overtraining of AIs — I can mislead any AI with one question.

So it starts producing absurdly unreasonable responses because the data training is very limited. So when we are talking about a complex disease like addiction — addiction is the number one most complex disease, because in addition to all of the biological complexity of a human being, it has also the complexity of the brain. And I'll give you the numbers — we have 6 billion nucleotides in our genome, in your recipe books. Letters in your recipe book. 26,000 genes, roughly 400,000 RNA transcripts.

And then you have billions and billions of neurons in all combinations — 100 billion neurons, in all combinations and connectivity. It's thousands of thousands of times more complex a disease than any biological condition we have ever known. Why am I saying this? More complex the disease — more complex the patterns of expression of a disease. More data do you need. You cannot get more data because everything is siloed.

That's why this is a culmination point of everything we have ever done in our lives. Justin — we are bringing the technology as a means of providing access to data. We are bringing AI, which is ready and mature but just doesn't have enough data — that's what we are suffering from, overtraining and hallucinations. We are bringing the expertise of people like you who understand the scope, understand the complexity of the disease. And we are going to merge everything — in this motile, intelligent way — and start doing research. We are not claiming...

So tomorrow we are going to issue a new type of drug or new type of protocol — we are still ignorant to it. Yes. We have seen some things which work well for some people. We have seen some things which do not work well for other people. What we need is to be scientists — pretty much connect all the tools and start learning, observing. To observe.

That's what it is — to observe first. To create a machine for observation. Because I am pretty sure — once a couple of years down the road, just like I gave the example of students and the NOAA data — better data. Yes. Give us two years observing data that is multifaceted, like we were talking about.

We will change the way addiction is even looked upon. The way addiction is even analyzed. Nobody does the same care about addiction research today as it is for cancer. And there are multiple reasons — stigma being one of them. And also, you know, the funding choices around addiction as well. I mean, people don't choose to have cancer. People make clear choices to become addicted.

And that's controversial. But I'll say it. Yeah. People... but then again, I mean, we are always pushed to the choices. I mean yeah. So it's not like you have full control or command of your choices.

1:37:33 Justin McMillen

This is the most important piece you just said. This is actually — and so, again, this country's idea that addiction is a choice. But the question is — and I talked to Charlie about this the other day on here — what things contribute to making that choice? And you just said it. How many billions of genetics and neurons... you tried to explain it to me one day — what contributes to making a choice to either drink this alcohol if I'm trying to stay sober, or not. Do I go to the right or the left?

And it's all of the things that happen behind that bottle — that are happening socially, neurochemically, genetically, biologically, mentally, psychologically, emotionally — all the variables that are dancing together that caused me to go right or left. And so we have to know everything about all these variables to such a degree that we can help people to always make a decision to go one direction.

And that equals disease remission — which is the successful target. We hit the target.

1:38:43 Vahan Simonyan

Absolutely. Yeah.

And one thing I want to say is — today when it comes to addiction, we are still in the Middle Ages. Middle Ages. Because we have never done — I don't want to diminish the role of very important scientific research. And you mentioned Nora Volkow — amazing woman, hundreds of publications. But they are individual spots of very high-level science. We haven't done fair treatment to addiction compared to other diseases like cancer.

Yes. Because we are still in the Middle Ages — because we don't have much data. We are ignorant. We are in the dark. Like, literally. You know, like appendicitis. That was called "side disease." It was prohibited — because the side pain — the left side. Is it left or right? Left side. So the challenge was — my religion was prohibited to open a human cadaver and see what's inside. Religions were prohibiting that. So "side disease" was treated — there were at least 500 recipes for treating side disease.

But people kept dying from it. Yes. And some things that were anti-inflammatory were helping at some point, but not really — until one guy, a Jewish doctor, found a dead body from side disease, opened it, and saw the ruptured organ in them. Yes. And that was the key moment. He was very brave. He did something for which he could have been punished.

He could have been punished for that almost. But then essentially what happened is the sultan — a Persian sultan, I think — got side disease and he saved him. That's how people started saying, okay, it's okay to look inside of the human body. It's okay to open — try to understand why the darn thing is happening in the first place.

1:40:24 Justin McMillen

I didn't know that. That's a really fascinating story.

1:40:24 Vahan Simonyan

It's okay to open — to do something like — try to understand why the darn thing is happening. I think the most complex organ is the human brain. As far as we know, in terms of complex connectivity, it's not the biggest but it's the most complex. So what we need to do is open the brain and get inside. And I don't mean physically open, of course — I mean, try to understand how it works.

Try to devise the machinery, the mechanism of action. And the brain doesn't work as a separate organ — it works with our biology, but it also works with our sociology, it works with psychology, it works with societies, with economics, with all of these things. So the only way you can understand that — to crack, proverbially crack it open — is by understanding the relations of our cognitive pathways with everything else that impacts our decisions.

If I'm going to drink this glass of alcohol or not — it's not alcohol, by the way. It's water.

1:41:42 Justin McMillen

Yeah, it's...

1:41:42 Vahan Simonyan

So the choices you said — like, yes, we admit addiction has a choice component to it. But is that truly a choice when you have a very complex organ which is out of balance due to a lot of external factors that we are pushed into — when we are a product of evolution? For millions of years our species survived because they made certain choices. The same choices, by the way, that made us sure up to today. So saying those are bad choices — I'm not sure if that's a bad choice. It's a choice that was appropriate given our evolutionary persuasion — in caves, in tribes, when the most probably alcohol from rotten fruits was the only antibiotic you could ever have, you know. So that choice is not a wrong choice.

It's an out-of-place choice.

1:42:32 Justin McMillen

Yes.

1:42:32 Vahan Simonyan

So I know when you say it's a choice, you don't criticize the people who make these choices, because you've devoted your life trying to save them. And I, we've discussed this. I know we are on the same side of this. But these choices are evolutionary choices. Yes. And people truly cannot fight the evolutionary choices. We can fight the conditions which make us put the wrong choice in front of us — in a society of access and abundance.

1:43:00 Justin McMillen

This is where you and I definitely get it. We get it. And I love this, guys, I love this — your thinking about this. And you know, because by the way — when we start understanding what causes people to choose to do things, and we look at it through an evolutionary lens, we start with addiction and we say — why do people make choices that continue to hurt themselves?

What are the contributing factors? How do we get people to make choices that could benefit them and help them? How do we set up the right environments for people to make good choices? Then it relates to everybody. Because there's nobody listening to this — whether they're struggling with addiction or not — who hasn't struggled with making decisions that are feeding indulgences and things that they shouldn't be doing, or they don't want to be doing.

So we all have a part of our mind where we can consciously observe our behaviors. And then you have the second part — the part of us that is doing in that moment. And we may say — for the sake of the future — right, because that's what discipline is, or that thinking is like — I don't want to eat this brownie today because I am trying to lose weight.

So I know that in the future I'm trying to go here — a huge part of our prefrontal cortex. Now we developed as a species — I'm going to make this choice today for some future situation. Everybody's struggling with that. And as you said — in the age of abundance, this is a subject that needs to be understood.

So where do you go to understand subjects? You go to the most extreme versions of them so you can really look at it. So why I love the idea of studying addiction — not only because it helps our life's work, but I think this is the subject of overindulgence and overconsumption and abundance. It's something that everyone is contending with.

And I think — I'm going to go really far and say that it's the future of our nation. Our nation is dependent upon us answering these kinds of questions, because we're destroying ourselves from the inside out by being consumption monsters, dopamine-chasing, gluttonous, overconsuming monsters. We're all overweight. We're all sick. Everybody's not doing well.

Not everybody, but a lot of people in this country. And now I think, as Americans — we have a sort of a duty to continue to show the world what's next. We've done that through all of history since we've been around. Now it's time for us to show how to be healthy.

But we have to consider so many things — again, you know, how do we do it? I mean, it's so — we've got on the table: we have data, we have access to data, we have computation, we have evolution, we have our species, we have where we're currently sitting in the societal modeling. And yeah.

1:45:23 Vahan Simonyan

And I think there is a fundamental challenge with addiction. If you really think about addiction — to what you'll see — every single element of addiction is about abundance and having too many choices and things.

1:45:23 Justin McMillen

Yes.

1:45:23 Vahan Simonyan

Talk more about that — the evolution part of it, and it being an evolutionary choice, because people have heard me say this kind of stuff in different ways but I'm nowhere near this.

But I learned about abundance from you. I mean, it just clicked in me — like that is obvious. How could I not even see this before? I mean, you pronounce the word like — the addictions are connected to abundance in society. Everything is too much of it. Let me give you this perspective. Like, why are we addicted to sugar?

Obviously that's one of the most important survival instincts, because it provides very cheap energy — a lot of it. And our ancestors who were roaming the savannas of Africa — from time to time they would find some sugary fruit from a tree and they would consume all of it, everything. Because then for two weeks they would go hungry anyway.

So they find a dead zebra they could steal from lions — or lions were just stuffed already — and they would eat it to the point of being completely immobile, unable to even move. Look — why are we getting fat here? But not in many other places. First, because fat is the part that goes up and down the least amount when you walk. So you don't spend as much energy carrying fat around your belt.

That's why we are designed to be creatures that gain as much energy as possible. That's why food addiction is here. Because literally everybody was hungry all of the time. And then — having genes that accumulate as much energy as possible were the best genes ever. Best genes. A person like you — so fit, so tuned to his physicality, acute feeling of his body, etc.

Imagine you going hungry for a month, or two, or eating some grass. That would be really punishing him. Somebody who had genes to accumulate more fat — they would actually be doing better. But unfortunately now we have 60,000-calorie refrigerators instead of dead zebras in our rooms. Everyone. Yes. Now imagine — you're not a hunter anymore.

Neither a hunter nor a gatherer. Just a very cold hunter-gatherer. The longest time evolutionarily — hunters and gatherers. Yes. Like, we only last — it's been like 10,000 years — we started controlling things around our environment. So now you take these hundreds of thousands of years of evolution as a hunter and gatherer. That means you work for days before you can find something meaningful to eat.

Yes. And then you eat it. Now we just eat it. And the work is from bedroom to the refrigerator. Problem. Big problem. The same with addiction to sex, the same — addiction to any other substance that we have. I mean, think about what was the role of substances that numb the feeling in your tongue and in your head or something.

You know, like — older people always had pain. We were evolutionary. There was no toothache dentist. Right. So what happens if your teeth are not good? You're in constant pain. Imagine the brain — imagine I divide people into two different families. One doesn't feel relief after eating — let's say cocaine leaves.

And the other one does. Who do you think suffers more? The one which doesn't have the cocaine addiction. You see — there's more suffering and more struggle, and most probably less success in hunting compared to people who die off. So basically you're saying that through evolution and natural selection, people are naturally going to be evolving towards addiction.

1:49:16 Justin McMillen

Towards addiction.

1:49:16 Vahan Simonyan

But we discussed this the other day. I mean — just to have sex, you need to: number one, be physically fit to be able to catch an animal or go gather nuts from trees or the best fruits that aren't rotten yet. All that stuff. So physical fitness is a factor for a man to find a mate. And then be strong — not be afraid of leopards — be able to protect the child.

All of these things. Yes. And then obviously, usually intermarriages would be between tribes. So let's say you are in a hunting mode and then you see a girl from the next tribe somewhere there. You would have to immediately react physically — be physically ready to perform the act. Yes. So the ability to see an image of a woman and be physically ready for mating...

You have only two or three minutes before the other tribe's males catch you and kill you. So being able to be aroused just from an image is already a big evolutionary benefit.

1:50:29 Justin McMillen

Yeah.

1:50:29 Vahan Simonyan

Because the alternative is — if you don't do it before they get to you, now the girl is gone and the tribesmen are trying to chase you or something like that.

And now we are asking the question — why are men addicted to pornography, for example. This is because our brains are stupid still. Because the moment we see the image, our bodies react, even though we understand those are just colored pictures.

1:50:52 Justin McMillen

That's so interesting.

1:50:52 Vahan Simonyan

Oh man. This. Yeah.

1:50:52 Justin McMillen

There's a... do you know that there was a study done on people in fantasies, and they did — they looked at men and pornography, and they looked at where women expressed something similar. And it was in erotic novels — they like to read stories. And then they discovered — and there's a hypothesis around this — that it has to do with them building a narrative in order for them to engage sexually. And the reason is because when they mate with one person, they're stuck with that situation for a longer time.

And so contemplation is much more important. And so women are always smarter.

1:51:34 Vahan Simonyan

Yeah.

1:51:34 Justin McMillen

Hey.

1:51:34 Vahan Simonyan

So yeah. So they're thinking about it. In order for a woman to be engaged, she has to constantly be creating in her mind this fantasy — like, this imagination of this person. Like — can I imagine them as the father of my children? Can I imagine them doing this?

1:51:56 Justin McMillen

Whereas men, it's like the silhouette and it's just go.

1:51:56 Vahan Simonyan

Yeah, yeah.

1:51:56 Justin McMillen

That constancy of behavior is important for a woman to raise a child. Clusters of mate behavior, all of that stuff.

1:51:56 Vahan Simonyan

Yeah. If you go to all addictions — if you truly think about it — a lot of addictions are all about some very useful construct evolutionarily.

1:52:25 Justin McMillen

Yes.

1:52:25 Vahan Simonyan

The abundance is the problem. Like we discussed multiple times — now all of that is so easy to get. But then the question is — yes, you and I will work hard and bring it up and start treating people based on their individual characteristics. But then at the same time, while the abundance of society is in place, there will always be some other struggle towards this.

1:52:56 Justin McMillen

Yeah, yeah. Well, what's good about this vision is — you know, you just spent this morning watching and observing some of our treatment model. And I think the key — and what I hope to see happen, and I don't care who does it, I just want as many people to do it as possible — is to consider addiction through this kind of lens, and then incorporate treatments that at least consider this, and try to bring some of these evolutionary ideas back into the forefront of treating addiction.

Because this mismatch — which you taught me about, between the environment and how we're genetically wired — we're wired for all these things you just discussed. But the environment changed. And so I believe everyone — no matter where we go with personalized medicine — I think most treatment models... and again, I'm biased, fully declared, I have a bias on what I do.

But I think based on this idea — this theory, which isn't so much a theory — I think it's true evolutionarily. We went through all of these things. We have to consider the whole picture, and we have to consider the condition of the human species. And that treatment should take all of those things into consideration. Which means that the base model will be something to do with making sure we honor all of these base things that are going away.

1:54:11 Vahan Simonyan

And then on top of that, you do personalized care.

1:54:11 Justin McMillen

And this is where it gets super interesting. But I wonder — I'll also wrap for time because, you know, you and I could literally do this for like two more hours. So I'm wondering — I love the discussion of science with you, because you make it not dry. You make it connected. You make it related to something to which you are so devoted.

1:54:32 Vahan Simonyan

And I saw you work. And I've seen you. I've talked to you. I've seen you thinking. It's amazing. I feel more connected to the science myself — being a scientist for like 40 years out of my 55 — I feel more connected to my own science when I talk to you. And that makes me proud of myself for being able to call myself your friend.

1:55:03 Justin McMillen

My friend? That's... that's important.

1:55:03 Vahan Simonyan

That's because people like you bring the personal touch to the science and make it about real things that can hurt. And you know — let me use the word "patient." It's really such a not-good word. Patient means like — hey, they can still wait more. Is that what it means? Patient is somebody smarter — anyway. What I saw today in your clinic — that older gentleman who was on the last day of his treatment.

And he's a completely different person today. He's somebody — his father who loves him, who takes care of him, and the kids are there. I mean, the word patient is wrong. When we say we'll treat "patients" — especially when we start showing numbers like 10,000, 100,000 — it completely loses the meaning. One person lost is tens of people suffering. Yes. I mean, that's somebody's daughter, somebody's son.

Some of this — somebody's mother, father, somebody, you know. So I really do appreciate — I saw these people. And you think you're treating these 50-odd men today — or I didn't count how many, actually. But in reality, you're treating their families. You're treating their future coworkers. You're treating their friends. You're treating their children. You're treating the entire community — socially, the impact.

I mean, sometimes when we — you and I — look at the insurance numbers and the payments and economic benefits, we need to do that. Obviously we need to do that just to understand are we doing well or not. But when you multiply the impact of that single dollar to what these people — how these people are changing — you cannot not take into account that what you are doing has to be repeated, has to be reproduced, has to be taken further.

1:56:52 Justin McMillen

It's not just ours now.

1:56:53 Vahan Simonyan

I'm hoping once again it works. And we do scientific models. If we take your methodology and make it what everybody else has to think is the way it should be done. And that's why I promise I will — once we do what we're going to do, once we go forward, we start looking at the numbers. We will make sure that everything we say is so scientifically rigorous nobody can criticize the numbers. But the eventual goal is not about numbers.

1:56:52 Justin McMillen

It's about these people.

1:57:09 Vahan Simonyan

Yeah.

1:57:09 Justin McMillen

It's their families.

1:57:09 Vahan Simonyan

That's why I love you, my friend. And you know, what you're doing is amazing.

1:57:15 Justin McMillen

I love you too. Thank you for letting me — for being here today. Being part of that. I really do appreciate that.

1:57:15 Vahan Simonyan

I'm so glad you came today. I'm so glad that you are my friend.

1:57:15 Justin McMillen

All right. Thank you. Cut it here. All right. Together like two hours. Okay. That's long. It was good. We could go five. We could go. We might want to consider doing, like, a CARES-IT-only one with very specific details.