Brainstorm AI Singapore 2024: The 1,000-Year-Old Human

  • 3 months ago
Christine Yuan HUANG, Founder and CEO, Quantum Life Alex ZHAVORONKOV, Founder and CEO, Insilico Medicine Moderator: Jeremy KAHN, FORTUNE
Transcript
00:00So, the title of this panel is deliberately a bit provocative,
00:03but there's several ways in which AI you could think
00:06about potentially extending human lifespans.
00:09And one is through better treatment of disease.
00:12That's one way of looking at it.
00:14Another might be through wellness, you know,
00:17sort of helping us live healthier lives.
00:19Another might be to actually try to treat aging specifically.
00:24So, taking some sort of compound
00:25that would reverse the aging process.
00:28And lastly, and probably most sort of sci-fi would be the idea
00:32that somehow we're going to, through brain-computer interfaces
00:35or some future technology, kind of merge
00:37with an artificial intelligence.
00:40And while our, you know, we might be able
00:41to extend our consciousness even
00:42if our physical bodies don't make it.
00:45I want to ask our guests, because they are experts
00:49in this area, where they think the most impact from AI will be.
00:53Christine, let's start with you and then Alex.
00:57Thank you so much.
00:59Thank you, Fortune.
01:00Thank you, Jeremy, for the wonderful panel.
01:01I'm Christine.
01:02I'm from Quantum Life.
01:04And the Quantum Life is actually doing quantum-enhanced artificial
01:07intelligence for precision medicine
01:10and for integrating human beings' longitudinal data to predict more
01:16and more accuracy plans for humans' health span.
01:21And, of course, maybe later on is lifespan.
01:24So we are so excited for this new technology merging up.
01:29And we are, we have already seen for humans' life
01:32and for the past like 30 years, we have already had a lot
01:36of changes in Medicare, in healthcare,
01:40in pharmaceutical, air-driven pharmaceutical development.
01:44But the question is, for human beings, we always think
01:49about birth, aging, disease, and death.
01:54We are in Singapore.
01:55Some people are talking in Chinese.
01:57We call it sheng lao bing, right?
02:00We never think about, we previously, we're all focusing
02:02on disease, which is disease and death.
02:07But how about aging before disease?
02:11That is how we wanted to introduce AI to improve before disease.
02:19And we have already seen during the past 10 years liquid biopsy,
02:23which is highly based on the bioinformatics being there,
02:30which is make great progress in cancer prediction
02:36and for disease prevention, which means a shift
02:39from the disease care before it had,
02:43based on the autonomous diagnostics,
02:46before it had to prevention.
02:49And next question that is going to be how we develop
02:54indications and interventions through the process.
02:57That is how AI will be introduced, and we are going
03:00to move forward for the next generation.
03:02Right. So this is on the idea of prevention as a way
03:05to forestall disease and to improve lifespan.
03:09Alex, I think you may have a, well, you may agree with some
03:11of that, but I'm curious what your view is on this.
03:13I think you might have a slightly different perspective
03:14on where you think the biggest impact will ultimately be
03:16on longevity.
03:18Sure. So I wear many, many hats.
03:22And one hat is in Silicon Medicine CEO, and another hat,
03:27I co-founded a conference called Aging Research
03:29for Drug Discovery.
03:31It's an annual event in Copenhagen every August
03:35for five days.
03:36A lot of people come to this conference, and we convene
03:40and talk about aging.
03:41So it's probably the largest event right now.
03:44And when I started the conference,
03:48most of the scientists were talking
03:51about diet, exercise, sleep, behavioral modification, right.
03:55And of course, you know, about 30% of the delegates
03:59at that conference are also going
04:01after very, very visionary things
04:03that are way beyond that.
04:05But at the end of the day, it always goes back
04:10to diet, exercise, sleep, and behavioral modifications.
04:14And what I'm afraid of the most is that in the next 10 years,
04:18I'm already 46, so kind of, for me, it's all like,
04:23you look into the future, it's all downhill.
04:25Sorry, for both males and females at that time.
04:29So whatever I do, it has to have more impact
04:32than just diet, exercise, sleep.
04:33So I think at most conferences, we need to start thinking about,
04:38you know, diet, exercise, sleep, prevention,
04:40diagnostics should be the baseline, right.
04:43So that's do what your mother told you.
04:45Just let's stop talking about us, do it, right.
04:47Some people are still smoking, right.
04:50So we need to have something much more substantial,
04:54a big breakthrough, a chat GPT moment in longevity in order
04:59to get people excited and get them, steer them away from,
05:03you know, diet, exercise, sleep, behavioral modifications,
05:06do what your mother told you.
05:07Because we know that works.
05:09It works to some extent.
05:10But billions of people live their lives.
05:14Many of them have optimized for diet, exercise, sleep.
05:19And they still aged, got diseases, and died.
05:22So we need to go radically to the next level.
05:25So far from what I've seen, nothing works in humans.
05:29There is nothing clinically proven to, no drug
05:34is clinically proven to demonstrate age
05:37reversal in human beings.
05:39So we are still trying to prove that some of the drugs that
05:43are known for 30, 40 years may have some impact,
05:47even though they work in mice really, really well.
05:49So like repamycin, for example, 15% to 20%
05:53consistent life extension in mice.
05:56And now people are doing combos, looking at different ways
05:59to different molecules to go after aging.
06:04But I think that we need to have a fundamental breakthrough
06:07in order to live longer.
06:09And of course, things like brain-to-computer interfaces,
06:14very promising.
06:15So there you're not only maybe reducing the rate of loss
06:20during aging, you might actually augment your function.
06:23Like, think about glasses, right?
06:25Very easy innovation, but it restored a useful function.
06:30Plus, nowadays with 3D glasses, you
06:33might be able, with augmented reality,
06:35you might gain new functions.
06:37So I think we need to go and look at the next level.
06:43I think that the next level is going
06:45to be drugs that target aging and disease at the same time.
06:48If you are just looking for using AI
06:51to discover drugs that only go after aging,
06:55I think you won't be able to develop it into a drug
06:59for many reasons.
07:00One is just basic economics.
07:01You won't be able to reimburse it.
07:03Most people misunderstand how difficult and costly
07:07it is to develop a drug.
07:09And also, you have to be committed for 10 years
07:11or longer, right?
07:12So I run in silico for 10 years.
07:15My most leading drug is in phase two clinical trials.
07:18So inshallah, as they say in the Middle East,
07:22if we are successful with phase two
07:23and we go and target the disease and show the proof of concept,
07:28in maybe like three, four years, we'll
07:30be able to show that it works on aging,
07:32hopefully, because that's how we discovered it,
07:34using aging research.
07:36So you have to be very committed for a very long time
07:38to demonstrate even a little bit of efficacy.
07:41Yeah, Alex, can you talk a little bit about how AI,
07:44though, is maybe speeding up some of that process?
07:47And I know you're getting this, but you
07:50have some of the most number of AI-discovered drugs
07:54currently in various phases of clinical trials,
07:56including in phase two a drug for idiopathic pulmonary
08:00fibrosis, which is a drug that's targeting both
08:03a disease but also has interesting implications
08:05for aging.
08:05So can you talk a little bit about, first of all,
08:07how AI can accelerate that process, that 10-year drug
08:11discovery process, but also this dual strategy
08:14of going after things that both target a known disease
08:17but may also have implications for age reversal?
08:20Sure.
08:21So first, I need to tell you, and that's probably
08:24useful for all of you, a story of how drugs
08:27are discovered and developed.
08:30So drug discovery and development
08:31is split in two phases, basically.
08:34Drug discovery, where you discover the drugs
08:37and ensure that you complete all the preclinical experiments.
08:41And then development, it's human clinical trials.
08:44Phase one, safety.
08:45Phase two, efficacy.
08:46Phase three, safety and efficacy.
08:48That part cannot be dramatically expedited,
08:51because you have to move with the speed of traffic.
08:54It's very regulated.
08:56But the part where you discover drugs
08:58can be substantially accelerated.
09:01And so far, we've demonstrated at least a 50% increase
09:05for the slowest of our drugs.
09:07In some cases, we've demonstrated
09:09that we can shave off 70% of the time and maybe 90%
09:14of the cost of drug discovery using AI.
09:18And drug discovery is split into three phases.
09:21One is formulating the mechanism of disease.
09:26So you need to understand why disease happens.
09:28So usually for Alzheimer's, for example,
09:30we still don't know why it happens and what's driving it.
09:34Then after you've formulated the mechanism of disease,
09:36you need to understand which proteins or other features
09:41in biology are driving those diseases.
09:43We call them protein targets.
09:45So you identify protein targets that may work in a disease.
09:49Then you design a small molecule, a drug,
09:52or a biologic, an antibody, that disables
09:56that kind of criminal that you've identified
09:58that is driving the disease.
10:00And that's a completely different kind of type of work
10:04that you need to do.
10:05Usually, that's two different separated areas
10:09in the pharmaceutical company.
10:11And then once you've designed the molecule,
10:13you need to conduct safety studies.
10:15So you need to go into animals and demonstrate, preferably
10:18in more than one animal, that the drug, well, first of all,
10:22works.
10:23So it's effective.
10:24And it's also safe.
10:25So you call it IND-enabling studies, safety studies.
10:28And then you can start humans.
10:30So in our case, we use kind of life models,
10:34the models that understand human biology from cradle
10:37to the grave and also understand diseases
10:41to pick those protein targets that
10:43are implicated in aging and disease at the same time.
10:45So there are a few.
10:46And then we have another form of generative AI
10:49that designs small molecules with the desired properties.
10:52Instead of looking for a needle in a haystack,
10:55you generate a bunch of perfect needles.
10:57And then you synthesize tests, go
10:59through the preclinical trials, go into clinical trials.
11:03And then we either take it ourself
11:05or we sell those molecules to other parties.
11:07So we've actually sold a couple of drugs very successfully.
11:11Last year, we did $1.5 billion in total deal value
11:15selling those drugs.
11:16So those AI-generated drugs that are faster and cheaper to make
11:19are actually of higher quality.
11:21So people are willing to pay a lot of money for it.
11:23Right.
11:24Christine, you were telling me as we were coming in
11:27that you're also sort of looking at this idea of creating a life
11:30model using AI, but maybe also to pick up
11:33on indicators of disease earlier.
11:37And then you can have earlier intervention
11:39is sort of your idea of how you could extend life.
11:42Can you talk a little bit more about that
11:43and what you're doing?
11:44And then I want to go to questions from the audience.
11:46So please think of your questions.
11:47Hopefully, we'll get one or two in.
11:48Yeah, exactly.
11:49Thank you, Jeremy.
11:50And just as Alex mentioned, because for drug discovery,
11:55it's a very, very long journey.
11:56And currently, we also have the drug discovery
12:00from the preclinical stage.
12:02And in silico, did amazing work driving forward
12:06the AI drug discovery to a clinical stage.
12:09But the question is, we still need
12:10to facing the clinical stage, right?
12:13We still need to take like 10 years
12:15to get through all of the regulatory approval,
12:18I have to say, for bringing in new medicine
12:22or bringing in effective drugs or cures or interventions
12:28to our life.
12:29It's not only about pharmaceutical.
12:31It's not only about drugs.
12:33It's also about one, regulatory.
12:36Secondly, what is a personalized right drug for ourselves?
12:41So during the past 30 years and during the medical,
12:44if we leave our Western medicine history
12:47like to the past 200 years, we will figure it out.
12:52OK, so actually, the medicine drug, the drug development,
12:56the system is all because of based
12:58on one fundamental for medicine, which is autonomy.
13:04And that is the fundamental curriculum in medical school.
13:07Like, I'm training as a medical doctor.
13:09So that is the first year curriculum.
13:14But during the past 10 years, we already
13:17shift our diagnostic window before the traditional autonomy
13:24because you think about it, how we diagnose a disease.
13:28We do autonomy.
13:30And we have our brain, our liver, our stomach, right?
13:34And now, what has happened?
13:36During the past 30 years, the next generation sequencing
13:40has already shown up.
13:42And a lot of molecular information
13:44has already been discovered.
13:46And together with the bioinformatics,
13:49which is the AI for the past 10 years,
13:53we have already developed several pre-diagnostic tools
13:58like Guerrero developed, the cancer preventions.
14:02OK, so the next era going to be, think
14:06about it, we have already shifted a little bit
14:08the cancer diagnostic stage from the late stage,
14:12like two and three, to the forward stage zero.
14:18And what is the right cure for the stage zero?
14:22And we need to develop a cure for the stage zero.
14:27And that is how quantum life wanted
14:29to discover from stage zero.
14:31And we compare that it's not only disease.
14:35We compare which is younger, healthy group and older group.
14:40And we discover what is the right intervention,
14:43what is the right target before any disease happening.
14:48And we develop cures for the stage zero.
14:51So that is how we can make drugs for extending
14:55the lifespan and healthspan.
14:58Then think about it.
15:00If we have this kind of interventions already,
15:04maybe drugs, maybe some other treatment like medical device
15:09can already be a treatment to the stage zero.
15:13And what is the next step?
15:15The next step, that is, how we can compare?
15:18Where is the right, if we are talking about the precision
15:23medicine, where is the database?
15:25How we can have the database?
15:27How we can make the individual personalized medicine
15:31for ourselves?
15:32So we also need to have the database for ourselves.
15:36That is the quantum life.
15:37You have to record a lot of data about each individual person
15:40in order to deliver the precision medicine.
15:42Because the data you have on the whole population
15:44may not be actually relevant.
15:45Exactly.
15:46That's interesting.
15:47Let's take some questions from the audience.
15:48At least let's get one in while we have time.
15:50Who has a question?
15:51Please raise your hand.
15:52We'll get a mic to you.
15:53I see there's a woman at the table over here.
15:55If we can get a mic to her, please.
15:57There's someone running over to you.
15:59And if you could please stand up and state your name,
16:01that would be great.
16:02And where you're from.
16:05My name is Miao Song.
16:06I'm Global CIO for it from GLP.
16:08So I have a question around the role AI plays in medical.
16:13And where do you see the obstacles
16:15of further utilization of AI, either in this not just
16:22precision medicine, but diagnosis of a disease,
16:24but also in a longer term health care,
16:27really care about people?
16:29Is that a compliance issue or government regulatory
16:32or other obstacles?
16:34So I guess, how do you make sure that AI is actually
16:38helping us bring out better treatments,
16:40maybe assisting doctors in a better way?
16:42And what are some of the hurdles to that,
16:43whether they're regulatory or not?
16:45I know Alex was talking about a number of regulatory hurdles
16:47that he's facing in terms of just getting treatments out.
16:50But I'm curious on this wider question,
16:51do you have a view on how AI might assist doctors
16:54in picking the best treatments for individuals?
16:56And then maybe Christine can answer.
16:58So sure.
16:58Actually, I will give the time to Christine on that.
17:01Because she is working in that area more than I do.
17:05So on a population level, how do you use AI for health care?
17:11That is the dream.
17:13It's actually the ultimate dream for human beings
17:16and for the current public health system.
17:19So I have to say that is how we wanted
17:22to bring Medicare beneficial to human beings
17:27and through medical doctor system as soon as possible.
17:32So first of all, AI for diagnostics,
17:34we can exactly enhance the diagnostic errors
17:40that make medical doctor radiologists to have
17:46a better diagnostics.
17:47And also, for medical doctors, can select better drugs
17:51and bringing the drug repurposing.
17:53Because some of the interventions
17:55may not need a new drug.
17:57We need to figure out how is the right drug for the right person.
18:01Maybe that drug is for traditional drugs,
18:04like metformin.
18:06Metformin is very cheap.
18:07And metformin, currently, we're also
18:09seeing the effect on extending human's lifespan.
18:14Basically, it's health span.
18:15But it's very cheap.
18:16So we are considering if we can use it and bring it
18:21to the health care system.
18:22Right.
18:22I have a question on this, which is sort of,
18:24I feel like there's a lot of these medical techniques that
18:26are sort of have been invented.
18:29There have been a number of breakthroughs.
18:30And yet, the distribution of that is a problem.
18:33So we have this big inequality issue,
18:34depending on where you live or even which doctor you happen
18:37to see, your doctor might know about this breakthrough
18:40treatment, might be able to prescribe it,
18:42might do genetic testing, might not do genetic testing.
18:44I feel like there's this great disparity in the health care
18:47we receive just on sort of these very arbitrary factors
18:50about where you live and what doctor you happen to have.
18:52I don't know if you can say anything about that, Christina.
18:54Maybe AI can help sort of bring the general standard of care
18:57up to a certain level.
18:58Exactly.
18:59Exactly.
19:00So yeah, I think that AI is the ultimate democratizer,
19:07enabler, and unifier, and equalizer.
19:11So now, with ChatGPT, everyone here in the room
19:15can have more access to knowledge
19:19than some of the best doctors just five years ago.
19:23So you can take a picture of your wound, for example,
19:27and let it annotate it.
19:29You have to hope it doesn't hallucinate some answer,
19:31though, Al.
19:31Well, yeah, it has a problem, right?
19:33But it has a very good demo of what
19:35is possible, because you give it a little bit more reinforcement
19:38learning.
19:39And if you don't need a snappy response,
19:41you can actually have a multi-agent system that
19:46will try to solve your question in maybe a little bit more time.
19:50Then it doesn't hallucinate, or you
19:53eliminate some of the areas where it could hallucinate.
19:58And it can answer very accurately.
20:00So it shows you that with technology,
20:03we actually not only equalize some of the professions,
20:09but we also make them very accessible to everybody,
20:12democratize, and we gain superpowers.
20:16And I think that we need to use this technology right now
20:20as much as possible to identify interventions that
20:23can target aging at its core and make this technology available
20:28to everybody.
20:28That's why we are very focused on small molecule drugs,
20:32because small molecules, unlike biologics,
20:34they are super cheap to make.
20:36Once they go off patent, they're available to everybody
20:38on the planet for peanuts.
20:41And they are also very easy to store, to transport.
20:44And people are very used to pills,
20:46rather than injecting expensive biologics.
20:49So there's lots of potential for AI
20:51to democratize access to health and access
20:54to, hopefully, these new medicines.
20:55We're out of time, everybody.
20:56But thank you very much for listening.
20:57And thank you to our panelists, Alex and Christine.
20:59Thank you so much.
21:00Thank you, Singapore.
21:01Thank you so much.

Recommended