- il y a 2 jours
Will the Next Medical Breakthrough be AI-Powered
Catégorie
🤖
TechnologieTranscription
00:00Sous-titrage Société Radio-Canada
00:30Sous-titrage Société Radio-Canada
01:06Sous-titrage Société Radio-Canada
01:09Sous-titrage Société Radio-Canada
01:13Sous-titrage Société Radio-Canada
01:16Sous-titrage Société Radio-Canada
01:20Sous-titrage Société Radio-Canada
01:26Sous-titrage Société Radio-Canada
01:29Sous-titrage Société Radio-Canada
01:32Sous-titrage Société Radio-Canada
01:35And yet, when we survey life sciences companies, pharmaceutical companies, only 5% of them say that they've managed to
01:44systematically embed AI into how they actually do science and into getting these medicines to patients.
01:52And that's because, as we'll hear, this is not just a tech problem. This is never just a tech problem.
01:57There are many ecosystem players that need to collaborate. There are many things that need to come together for this
02:02to work. And so that's what we're going to dig into a bit today.
02:06So maybe I'll start with a question. The panelists here are from a range of very diverse areas of the
02:12ecosystem. Maybe you can all talk about how you think AI can and will contribute to scientific advancement and patient
02:21outcomes. And maybe, Marion, we start with yourself.
02:23Yeah, sure. So thanks for this question. There are so many exciting things that AI will enable. And I think
02:33one of the first things that will be improved, and it's already happening, is improving patient diagnosis.
02:40Because AI will and is already enabling to analyze so many data that the human brain was not able to
02:50analyze alone. And now putting together all the modalities, imaging, and biological data, clinical data, all together, is improving our
02:59knowledge of the disease, how to diagnose those diseases, and therefore will improve how we treat those diseases.
03:06I think also it will enable better devices to monitor how the patient is evolving under the drug, to switch
03:14the drug if needed. So many, many different fields of application. It's already happening, but it's the very start of
03:24it. And I will let my other co-speakers give their opinion about it.
03:29Thanks, Marion. I can only second your excitement about this. And I'm just catching my breath having run from a
03:36late train from London. I hope somebody in this building is working on trains while we worry about health care.
03:42I work at Isomorphic Labs. And our goal is to basically change how drug discovery is done with artificial intelligence.
03:51And I think there's a huge amount of opportunity in this space, of course.
03:56Drug discovery deals with understanding diseases. It deals with designing molecules that are going to help us cure patients.
04:04And currently, the way this process is carried out is in a fairly artisanal way. We have folks with a
04:12lot of experience in chemistry and biology who are using a lot of their intuition and experiment in order to
04:18be able to generate and validate hypotheses.
04:21And there's a huge opportunity for us to actually take this experimental work into the computer and be able to
04:28build models that are going to allow us to design in a very rational manner.
04:33And I think this really gives us an opportunity to get rid of massive amounts of churn, massive amounts of
04:39failure that is really common in the drug discovery space because we're going to be able to really reason about
04:45this space in a rational, in a mathematical way.
04:48Over to you, Julia.
04:48Thank you. Hi, everyone. As the investor on the panel, my perspective absolutely echoes what my fellow panelists have said.
04:57You know, we're living in a world where we're living longer, but spending more time in ill health.
05:04And where you live really plays a huge factor in determining your health outcomes.
05:10So, using AI to improve health outcomes for all patients is what we're looking to back, whether it's earlier diagnosis,
05:20better drugs faster, the right pre-treatment for the right patient.
05:24Absolutely agree with those.
05:26Another area that I'm finding hugely exciting is in clinical trial space and using in-silico data.
05:33So, computational models, digital twins to generate data faster within a clinical trial.
05:42That's an incredible spread of opportunity, you know, from drug discovery, reducing the trial and error, diagnosis, devices, better equality
05:51and healthcare provision, clinical trials.
05:53With all that excitement, I think we would be remiss if we didn't talk about the difficulty of doing these
06:00things and how hard this is.
06:01Maybe you could each share what you think are the biggest potential pitfalls, the biggest risks with this area.
06:08Maybe, Sergei, we could start with yourself.
06:09Yeah, thanks, Alex.
06:11I think there are a few things to be aware of.
06:13I mean, in the end, biology is a very complex system and it's a system that evolves.
06:19And I think it would be foolish to think that we could solve, you know, disease in its entirety.
06:25You know, what we really need to be prepared for, I think, is a long, long battle in a sense
06:30and arm ourselves with the best tools in that battle.
06:33Because when you think about infectious disease, when you think about diseases like cancer, you know, there's constant evolution going
06:40on.
06:40And so, you need to be able to have answers to how that complexity evolves.
06:44And so, you know, I think one of the pitfalls is to assume that AI is going to be this
06:48magic silver bullet that is, you know, going to cure cancer in one fell swoop, for example.
06:53And so, to me, this is one of the things that's quite important to be cognizant of.
06:57The other is the quality process.
07:00When we deal with healthcare, of course, it's important for us to understand the safety and the efficacy of the
07:06treatments that we're designing.
07:08And so, I want to be sure that even though we're designing AI systems to help us design better medicines,
07:15we don't get ahead of ourselves in assuming that we can, for example, just put these straight out to market.
07:20And I'm really actually grateful for having, you know, very thorough mechanisms that exist in the industry for us to
07:26ascertain the safety and efficacy of these medicines.
07:29You know, we have clinical trials and they're quite long and expensive and error-prone and failure-prone.
07:35But actually, there's also a lot of opportunity to contribute there from a machine learning modeling perspective.
07:40So, I think we want to make sure that we have that in place.
07:43We want to make sure that we're making that as efficient as possible.
07:47But I do hope that in the future, you know, to me, machine learning is mathematics.
07:52And one of my dreams is that in the future, we should be able to actually mathematically prove some outcomes.
07:57As our models get better and better, we should be able to reason and say, this is going to work
08:02and then demonstrate that it is.
08:03And I think that'll be a really great world to be living in.
08:07And from my perspective, I think there are essentially two structural challenges that we should be considering.
08:15One, I think health systems are actually not really set up to diagnose.
08:21We're really set up to treat.
08:23And that just means that we need a paradigm shift in order to ensure both adoption as well as the
08:30commercial success of these amazing technologies that are now being able to come to market.
08:35I think that the second part is actually on the investor side.
08:41Their VCs, most VCs are actually not set up to take on the risk associated with the longer lead times
08:48that are often associated with developing diagnostics or therapeutics.
08:52And so, we need, in my opinion, more specialist investors.
08:55And I think there's a really important role for policy and government to play in supporting these companies that are
09:02tackling some of these biggest challenges.
09:07All of that is so interesting.
09:09I think what I can add to your great comments is that we also need to be aware that the
09:16world might not be fully ready.
09:17The technology might be, but we need to also switch our mindset and really be aware of what the AI
09:26can and cannot do and then really help us help the world, the authorities who regulate.
09:34So, it's also a challenge.
09:36And as you were saying, Sergei, biology is moving.
09:41It's something that we need to adapt to.
09:45The parameters that we are monitoring and using are moving.
09:49And so, sometimes the model is designed for a specific task and cannot apply to a different task.
09:55And we are seeing that with diagnostic tools that are doing great because they've been trained on a very, well,
10:02specific data set.
10:03And it's not applicable everywhere.
10:05So, we need to collect as many data as possible so that the models can see as many possible varied
10:12data as we can.
10:14So, well, trust, be prepared to use AI.
10:20And so, our mindset to everyone, the patient, the doctors, the authorities, everyone.
10:27So, technical risks, ecosystem, structural risks, people, mindset risks, lots for us to be cognizant of, you know, buyer beware
10:35as we step into this space.
10:36When you think about your corners of the ecosystem, what are the kind of challenges do you see that need
10:42to be overcome to actually harness AI, to actually pull this through to impact?
10:46And how do you think we'll go about doing that?
10:48Maybe, Sergei, we start with you.
10:50Yeah, thank you.
10:51It's hard to pin a single thing, but I want to mention two, and the two are data and compute,
10:58which may be not that original of a thought, but I do want to expand on that thinking.
11:03Of course, all machine learning models need data.
11:06And, you know, a lot of what we read about, about machine learning in the news and in the wider
11:12sort of publication space is around text models or image models.
11:17And these models have the whole of the internet as training data and is intuitively clear to us when we
11:24see text or an image what that means usually.
11:27When we deal with biological data, it is not so clear, actually.
11:32It's usually representing something quite complex that is going on at a molecular level.
11:37It's usually measuring something experimentally with a degree of error, and it's usually quite expensive to produce this data.
11:44And so when we think about how to build these systems, it's really important for us to be able to
11:50collect data sets that are going to allow us to actually build general systems.
11:54You know, this has been one of the shifts in machine learning over the last 20 years.
11:59We used to build models that were very specific to a data set that described a particular narrow problem in
12:05a space.
12:06And as we've seen in the wider machine learning space, you can build very general models, and then you can
12:10use them to solve many different tasks.
12:12And so I hope to be able to also be in this regime, in the healthcare, in the sort of
12:17drug design space, where we can build general models using large data sets.
12:22And I think, you know, in that space, it's very important for us to also work together.
12:27You know, I think humanity can give itself a pat on the back for creating really incredible public resources, like
12:33the Protein Data Bank, for example, which made AlphaFold possible, which has been developed by Google DeepMind.
12:39And the latest version, AlphaFold 3, by Biosomorphic Labs and Google DeepMind together.
12:44And, you know, then released over 200 million predictions of protein structures and placed them in AlphaFold DB together with
12:52European Molecular Biology Laboratories.
12:54So these resources are really, really important.
12:57And I think we need more of them to be able to create these systems.
13:00And the second part of what I think the challenges are is around compute, because actually training these large general
13:07models, it's very, very expensive.
13:09It can cost tens or hundreds of millions of dollars to actually train fully a large-scale system like this.
13:16And this infrastructure, you know, is quite rare.
13:20We're quite lucky, for example, at ISO labs, being part of this Alphabet ecosystem to have access to large-scale
13:25infrastructure,
13:26which means machine learning researchers have a lot of freedom to explore different ideas.
13:31Machine learning is a very empirical science.
13:33You need to try many different things.
13:35And so that creativity is unleashed by having access to this infrastructure.
13:39But I see, you know, more widely, actually, a dearth of access to this type of infrastructure,
13:44especially so potentially in academia, where it's not uncommon for universities to have a fairly small-scale access to compute.
13:53And I think this is a problem we need to think about more broadly as a society,
13:56because, of course, we're expecting our universities to be at the forefront of research.
14:01And, you know, there's a risk that they may fall behind.
14:06If I can add to all that, I totally agree with all you've just said, and it's very technical.
14:13But there are also some challenges around regulations.
14:17Julia, you were mentioning digital twins.
14:20And, yes, we are so all excited about being able to use what is a digital twin,
14:26so to avoid having to enroll patients in the clinical trial
14:30and ask AI to simulate what could happen in that patient.
14:35And it's going to be easier for patients, faster clinical trials.
14:41But then will the authorities approve that?
14:45I'm not sure they are ready yet.
14:46And this is a very important challenge we need to face.
14:50There are many, well, 8,000, I don't know how many devices that are FDA-approved,
14:56but it's just a device.
14:58When it gets to giving the approval of a drug based on what a model said, it's something else.
15:04So that's one of the challenges I think is very important to keep in mind.
15:07And there is another challenge that I think is important.
15:13I was mentioning trust.
15:15But then I think we need to communicate more.
15:18This is also one of the challenges, because some people might be scared.
15:23Okay, it's an AI who predicted which drug I should take.
15:26And they might be scared about it, but AI will help doctors, will help physicians, biologists, data scientists.
15:37But the decision will remain human-based.
15:40It will be a tool that we will be using to help us identify maybe a drug we wouldn't have
15:46thought about,
15:47because there is an ongoing clinical trial somewhere, and we were not aware about it.
15:51So there are challenges on the user side and on the authorities side about, well, I think,
15:59explaining exactly what is AI doing and what it is not doing.
16:04I couldn't agree more.
16:06And just to build on that, I guess in the case of early detection,
16:10I completely agree that one of the biggest pitfalls is understanding human behavior
16:16and understanding psychology and what motivates an individual to take action to improve their health.
16:24And so exactly that, Mary.
16:26And I think communication and the incorporation of social sciences within teams
16:32who are bringing these amazing technologies to market is absolutely fundamental.
16:37So being able to explain the value of whether it's a treatment or a new technology
16:44and expecting the person to take action based on that is definitely something to consider.
16:50That was an unexpected emotional rollercoaster of a question from, you know,
16:54a pat on the back for humanity to fear and then back to social sciences.
16:59All probably necessary ingredients in making this work.
17:01You all mentioned different aspects of the ecosystem, if we'll call it that.
17:06Academia, regulators, you know, all these things.
17:10You are all representing here different parts of that ecosystem.
17:14How do you think about the most important points of collaboration,
17:18where that works and where we can do better?
17:21Maybe, Julia, start with yourself.
17:22Yeah, absolutely.
17:23I mean, I can share a little bit about how we've thought about building community around health care.
17:30And it's actually been based on two pillars.
17:32And maybe it's slightly counterintuitive.
17:34But the first part is actually being place-based.
17:38So we are, as an investor, we're based in Summerstown in Camden.
17:43It's actually the poorest part of Camden in London.
17:45It's got the highest child poverty rate.
17:48If you live in Summerstown, you live 15 years less than if you live 15 minutes up the road.
17:53So the public health aspects just on our doorstep are very, very pertinent and relevant to us.
18:00But just across the road from where we're based is the CRIC, one of the largest bioinformatics centers in Europe,
18:07around the corner from Isomorphic Labs, from DeepMind.
18:10Also on the doorstep is Great Ormond Street Hospital, other hospital systems.
18:15And so the way that we've built a community has actually been place-based.
18:19And we've had amazing individuals who have joined us for community-driven events that have surfaced various topics,
18:27from how to bring treatments to patients and how to talk about AI in health care.
18:34And because it's place-based, it has just meant that we can surface and discuss conversations in a very open
18:41forum.
18:43And then I guess that has just led to, so non, yeah, basically just conversations that have been non-transactional.
18:51And I think from that perspective, you have, as you said, all of the different stakeholders
18:56who are able to speak and build relationships, which enables us to move forward in conversations.
19:04Sergey?
19:06Yeah, I work in this drug discovery space, and the space is absolutely vast.
19:11And going at it alone is really hopeless, I would say.
19:16And so I think it's very, very important to partner.
19:19And when one thinks about what it takes to actually go from a disease to a medicine that is going
19:25to help patients,
19:26it's a very long and arduous journey.
19:28So, you know, in numbers, I guess, our current statistics are such that it takes about 13 years on average
19:34to bring a drug to market from the beginning of the research program to when it's on the market.
19:39It costs $3 billion on average.
19:42We go through most of the process to enter clinical trials, put our best candidates forward,
19:47and 90% of those candidates fail in clinical trials.
19:51So there's huge amounts left to be desired.
19:55There's huge amounts of opportunity.
19:56And so when you go after it, you need to do what you're good at,
20:00and you should find partners that are going to help you do what you're not.
20:03And, for example, the way we think about it at Isomorphic Labs is we have a lot of strengths in
20:09modeling chemistry,
20:11in modeling certain basic areas of biology,
20:14but we have no idea how to manufacture medicines, for example, or how to do clinical trials.
20:20And so we've partnered, for example, with Eli Lilly and with Novartis that are two sort of top-tier pharma
20:26companies
20:27that are going to help us do that and be more successful together.
20:30And I think as part of doing that, it assumes a degree of openness.
20:34It assumes a degree of exchange of information, of best practices.
20:38And actually, this has been a really, you know, really great experience for us.
20:41And so I'm really looking forward to the medicines that we're designing now hopefully making their way to patients over
20:48the coming years.
20:49Yeah, well, I couldn't agree more being in the world of drug discovery at Sanofi.
20:55But my background is I'm a pathologist.
20:58I used to work at Gustave Rousset Institute.
21:01I also am from the side of diagnose better patients and use the drugs that actually pharma industry is developing.
21:07But we need to collaborate.
21:09And the collaborative mindset is crucial because what is important to build AI models is data.
21:16And if we're not able to share our science, our data altogether, we'll never have, well, the strength and the
21:24power to really impact as we could if we were able to share.
21:29So being able to collaborate is mandatory.
21:32And especially since exactly what you said, there is so much expertise in very specific tasks.
21:40So you are very good at one thing and your colleague is very good at developing the model.
21:46And then another colleague will be able to generate the actual molecule.
21:50And if we don't talk to each other, if we are not able to share in a very good mindset
21:55all that information,
21:56then the drug will never penetrate the market, never will be delivered to the patients.
22:01So the collaborative mindset is very important.
22:05And also about academia on the diagnosis perspective.
22:09I mean, if we want to improve how early we can diagnose the diseases, we need to work with hospitals.
22:16They have the real world evidence.
22:18They have the data.
22:19And so patients must also share consent at some point to anonymously share some of their data if we want
22:26to improve our data collection to apply to our models.
22:31So it's a real mindset change that we need to apply to be able to work all together for the
22:39same goal.
22:42Lots to do and almost it's inevitable that we'll have to do this, right?
22:47To your point, we need specialization in different skill sets.
22:50We need people who are extremely deep in their aspect of it.
22:53And therefore, we need to partner together.
22:56Maybe let's try and end where we started in terms of inspirational, forward-looking, what we think this will do.
23:02Take us 10 years into the future.
23:04How will the medical treatment landscape have evolved through AI?
23:10How will we be looking at a fundamentally different landscape?
23:13Maybe whoever has the strongest opinion, speak first.
23:18Go on, Junior.
23:19Yeah, sure.
23:19I guess I have two predictions.
23:23And one is just around, I think that we will actually be routinely screening for cancers at home non-invasively.
23:32And I believe that that is a world that is coming together and will be a reality in 10 years'
23:38time.
23:39I also think that the world's most valuable company will be a healthcare company.
23:44I think that that's completely also within our reach.
23:47We know that Nova Nordisk is Europe's most valuable company at this point in time.
23:52And so it's possible for the world's most valuable company to be a healthcare one.
23:56And I hope that's the case.
23:58Great.
24:01I'm a real technology optimist.
24:03I've been building technology since probably the age of 10.
24:06And, you know, I feel like we are building the future of healthcare with machine learning now.
24:12At the same time, I want to stay grounded and keep things just a little bit out of the realm
24:18of science fiction.
24:19Although, I do have a science fiction-like dream as well.
24:22But I do think it's possible for us to build really, really accurate models of chemistry and of very basic
24:30things in biology that are just going to allow us to do the types of research that is required to
24:36bring new medicines to patients in a completely different way.
24:40You know, all those years that we spend doing experiments and validating and failing and really coming against the edge
24:47of human understanding of these mechanisms, I think we're going to be able to overcome.
24:53And we're seeing evidence of this with what we've already been able to do.
24:57So, for example, you know, only a couple of weeks ago, we released this new model called AlphaFold3 that we
25:03had co-developed together with Google DeepMind.
25:06And so, while only three or four years ago, AlphaFold2 was this big model that, you know, quite thoroughly solved
25:13the problem of predicting protein structure.
25:15How will proteins fold given an amino acid sequence of this protein?
25:19We've been able to actually extend this and start solving the problem of predicting how do different molecules interact with
25:26these proteins?
25:26How do proteins interact with DNA or with RNA?
25:30And so, these become building blocks of these really, really general machine learning systems.
25:35And I think we're going to see this kind of evolution of foundational models in the life science space.
25:43We see this already happening in the natural language space with models like Gemini or GPT.
25:49We see this with models like Imogen or DALI where you create these general models and then they give rise
25:56to a whole ecosystem.
25:57You know, you can create hundreds of startups and companies and products.
26:00And I think we're going to see something similar take shape in this space as we create this kind of
26:06foundational layer of reasoning about chemistry and biology.
26:09And, you know, one of our missions at our company is to be building that foundational layer now.
26:13And so, I'm really excited for this future where 10 years from now, we're going to be able to be
26:19operating in the space where a lot of this experimentation happens in the computer.
26:23But my sort of, what I think is the sci-fi dream a little bit is going away from this
26:29notion of medicine as a way to treat disease and go much further towards prevention.
26:34You know, if you're fans of sort of Peter Attia or other thinkers along those lines, I think it's really
26:40important because, of course, we're quite good at treating diseases.
26:43But we know that the older you get, the more disease burden there is and you survive, but actually your
26:48quality of life is not that great.
26:50And so, my hope is that we're going to be able to use these systems to be able to actually
26:55overcome that.
26:56And I think that rests upon what you said with regards to diagnosis.
27:00I think it rests upon what you said with regards to being able to do a lot of this screening
27:05in your house.
27:06You want to have the background information about the people's genomes, about the microbiome, about their daily biomarkers accessible in
27:15your home.
27:16You don't want to be going to the hospital or to a clinic, you know, every few days to check
27:20in that you're okay.
27:21And so, I think there's going to be this explosion in devices and our ability to actually make sense of
27:26this data.
27:27And we don't, you know, we need to not lose track, of course, of sort of the sensitivities of that,
27:32of making sure that that data is safe and is handled in an ethical matter.
27:36But to me, that's really kind of a future I'm really looking forward to in 10 years.
27:42I'll be pragmatic.
27:45I remember when I started my medical training, it was really when next generation sequencing, it was at the top
27:57of everything.
27:58And we were all saying microscopes and pathologists will just disappear thanks to molecular biology.
28:04We are still there, 20 years later.
28:08So, I just want to say, yes, I'm so excited about those technologies.
28:15Yes, it's evolving very fast.
28:17But I think in 10 years, we will get used to having AI tools to help us in our daily
28:25practice in hospitals for patients also to track and monitor their diseases at home.
28:31Like for diabetes, it's already happening.
28:33And it's really so much more comfortable for patients.
28:36So, we will be at the start of using concretely those AI tools on a daily basis.
28:43But I don't know if we will be, in 10 years' time, ready to apply all those models everywhere.
28:50But still, I'm pretty sure a lot of progress will be made until then.
28:54And most importantly, we will get used to using AI as a tool to do better, to understand better, to
29:02track better, to make better decisions.
29:05That's what I feel is going to happen.
29:08So, a brave new world in which we'll all hopefully still have roles to play and not supplanted by AI.
29:14We have a few minutes, so I might cheekily ask a bonus question, my prerogative.
29:18Anyone in the audience who's wrestling with all the challenges we've just spoken about and the scale of the opportunity
29:24is so big,
29:25what advice do you have for people who are thinking about this in their own companies?
29:29How do they capture the value?
29:31How do they go about this?
29:32What guidance would you have for the audience?
29:36I can start.
29:37I guess my biggest advice is to get informed, to get curious.
29:42This has been also, you know, I've been on a learning journey myself in the sense that I started my
29:48career as a computer scientist
29:49and worked for about 10 to 12 years in various technology industries that had nothing to do with healthcare, in
29:56fact.
29:56And then, at some point, actually, right around the time that you described, Mahion, when Next Generation Sequencing came around,
30:04I actually, you know, really discovered this passion and I realized that we can use technology to really, you know,
30:10to really help advance healthcare.
30:12And, you know, some of the things that really helped me were, you know, some of the online resources that
30:17were available, you know, massive online courses, you know, people's podcasts.
30:21And I feel like I just dove in headfirst into the space and then have never really looked back.
30:28And so, for anyone who is interested, I just recommend to get really curious about the space.
30:33And there's so many threads to follow, you know, whether you do that on Twitter or on podcasts or however
30:39you do it and at the level that you're comfortable with.
30:42But it's, you know, one of the most beautiful areas because it's so relevant to all of us.
30:47Healthcare is, but it's also got so much beauty.
30:50You know, we're literally thinking about the secrets of life and so it's really incredible.
30:54I would just add to that and we're having some really amazing conversations with AI researchers who are leaving really
31:04amazing institutions and they want to start companies in and around AI for healthcare.
31:09And so, you have engineers who have this amazing grasp of data and engineering, but then need to get themselves
31:17educated on the healthcare market.
31:19And so, what I do is I say, well, welcome, welcome and like exactly as Sergey said, get yourself informed.
31:28I will connect you to various people in the industry so that you can learn whether it's access what data
31:34sets are available to you, whether it's patient groups that are able to talk to you, whether it's clinicians that
31:40are open to collaborating with you.
31:43I think having a customer-focused lens on essentially the kinds of products that you're building, I think is incredibly
31:52important.
31:53But yeah, for anyone who's interested, I would just say welcome and yeah, here to support you.
32:00Yeah, I will. I just, of course, agree with all you just said.
32:05But I think also, if you are interested in the field, if you are passionate in the field, there is
32:10so much space for everyone.
32:12It's just the beginning of that revolution.
32:15We are just at the beginning of the revolution and so there is space for so many different actors.
32:20Be curious, meet people and I think if you want to work in that space, you will.
32:27Like, I'm an MD. I started in a lab and now here I am.
32:31Because I was interested in the field, I was passionate about understanding, wow, those models can do the data like
32:39I do it.
32:39I want to understand. I want to be in that field.
32:42And so there is a lot of interest. There are many topics to be covered.
32:45So we need all the people who are interested in the field to work all together if we want to
32:50make huge progresses, I guess.
32:53Perfect.
32:54Perfect. As expected, a rich discussion.
32:58And there it is. If you want to live in a world without cancer, if you want to live in
33:02a world where all these things happen at home,
33:04get educated, become an expert in AI, become an expert in healthcare, become an expert in pathology.
33:09There's plenty of space for everybody.
33:11Please join me in thanking a wonderful panel for a wonderful discussion.
33:16Thank you.
33:30Thank you.
Commentaires