Moderna, Google DeepMind, Biocon on AI in Science

Author: Bloomberg Television · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Moderna, Google DeepMind, Biocon on AI in Science · Channel: Bloomberg Television · 30:20 · Watch on YouTube

Topics: AI governance and regulation, Health and science

Summary

Leaders from biotechnology and AI discuss how artificial intelligence could transform scientific discovery, drug development, agriculture, and personalized medicine. They highlight opportunities to connect biological data, accelerate research, improve clinical trials, and potentially make medicines more affordable, while stressing that human biology remains complex and timelines should not be overstated. The panel also considers public trust in science, the need to involve communities in AI projects, and whether advanced AI agents require regulation. They conclude that the immediate priority is using AI well to improve current work.

Key points

AI may change how scientific discovery is conducted and help researchers explore patterns beyond what human imagination alone can find.
Combining AI with biological understanding could support medicines that address networks and systems rather than single targets.
AlphaFold illustrates how AI tools can make protein-structure predictions accessible to researchers and help connect existing medicines with new applications.
AI could shorten drug-development timelines and help reduce costs, but clinical trials remain a major challenge.
Near-term clinical applications may include improving trial participant selection and reducing false positives in cancer screening.
AI-supported crop research and farmer-facing tools could help address agricultural challenges in India and other regions.
Personalized medicine may benefit from combining genomic information, biological models, and ongoing experimentation.
Trust depends on the scientific method, clear communication, and meaningful involvement of communities and practitioners.
The panel discusses regulating commercialized AI agents, distinguishing their intentional behavior from the capabilities of large language models.
The immediate focus should be on applying AI effectively to make a difference in existing work.

Chapters

Start	Chapter	Summary
00:02	AI and the future of scientific discovery	The panel considers how AI could change scientific discovery and how machine intelligence can help researchers understand nature.
04:58	AI tools, drug development, and affordability	The speakers discuss AlphaFold, drug-development timelines, clinical trials, and the prospect of getting medicines to patients faster and at lower cost.
11:57	AI applications in agriculture	The discussion turns to sorghum research in India and AI-powered tools designed to support farmers.
15:15	Personalized medicine and scientific trust	The panel explores how AI and biological intelligence could advance personalized medicine, alongside concerns about public trust in science.
21:42	Communities, communication, and human skills	The speakers call for clearer science communication, community involvement, and continued development of human creativity and curiosity.
24:20	Biotech opportunities and AI regulation	The panel considers biotech innovation in India and debates how commercial AI agents and potential unintended consequences should be regulated.
29:14	Safety, affordability, and closing remarks	The speakers summarize work on AI safety and conclude that the immediate task is using AI effectively in current scientific work.

Quotes

I think the answer is everything. — 00:36

if you can compress that time, there is of course going to be a cost element to the use of AI itself — 07:39

we have to be more humble. — 11:24

LLM’s are capability. Agents are intentionality. — 27:31

the main emphasis today is really about how best to use AI to make a difference to what we’re doing today. — 30:20

Concepts and entities

AI, AlphaFold, clinical trials, personalized medicine, AI-powered platform, agents, Moderna, Google DeepMind, Biocon, Nubar, Laya, Kieran

Full transcript (original language)

00:02 It’s very good music. Hi everyone. I have the best job in the world because I have three wonderful leaders to try and understand what they believe will change what is still holding us back and who will benefit from this next next era of discovery. Brings together three perspectives. Newar on creating new companies and scientific platforms. Laya and building AI capabilities for science and Kieran on biotechnology drug development and bringing innovation to patients. So, thank you all for joining us. Noar, let me start off with you. What do you think in your wildest dreams AI will help us discover that we can’t

00:36 even imagine yet? We can’t imagine, but you can. >> It’s a tricky question because I think the answer is everything. So, in other words, I think how we discover everything in the future will be fundamentally different and it will not be limited by human imagination. Um, in nature, nothing is discovered through human imagination. It just exists. And now we’re going to be able to do that in in multiples of what humans have been able to do. And it’s an exciting but also scary prospect. >> LA, do you think AI is accelerating science or actually just fundamentally

01:09 changing it? >> I think it’s helping us understand the world in a way that we never imagined. As Nubar said, much like a microscope of being able to go very narrow and deep or a telescope to go out wide. And I do think that maybe the limitation is our ability to ask creative questions to be bold enough and have the courage to ask. What do you think is is the next AI breakthrough? >> Well, it’s definitely going to be very different to what we’re doing today because right now, you know, biology is

01:41 a very empirical science and I think AI will enable us to do it very differently. It’ll allow us to collect connect very unrelated dots. So I’m looking at developing and innovating new medicines that are not about targeting single targets. You know, it’s not those single targeting drugs, but I really believe that you’re going to be able to with a combination of understanding biology and leveraging AI, you’re going to be able to basically holistically

02:12 look at networks, signaling channels, and neural circuits. And I think that’s the way you’re going to develop new drugs, you know, multi-specific, multi- faceted drugs. >> But Nubar, is there something that actually remains, you know, intrinsically human in scientific discovery? >> Um, well, maybe maybe I’ll reframe it in this way. So I I know that we’re at an era where we’re trying to put the human back into life and existence because it’s being it’s being challenged by this other life form that we’ve created, at least thinking form. I do think that how

02:46 it is to be human is going to change in the scientific fields in the following way. I think we’re today obsessing over machine intelligence. We can call it super intelligence, AI, whatever. But I just call it machine intelligence because it’s a machine form of intelligence uh such as it is. Then we have human intelligence. But I think what we have failed to realize is that there’s another form of intelligence surrounding us which is nature’s forms of intelligence. Every living cell, every cell that processes information is also a form of intelligence. I think once we realize that, we’re going to realize that what biology has been, which is humans trying to understand

03:19 nature’s intelligence, is now being supplemented by humans trying to understand machine intelligence. And we’re not going to be good at either one in my view. And we’re going to have to figure out how to use machine intelligence to better understand nature’s intelligence, whether that’s the immune system, whether that’s viruses. I mean, people often debate with me what how I could say that a virus is intelligent. Well, a virus that could take down most of the immune system on the planet through a pandemic. If that’s not intelligent than then our bodies must be pretty weak. The fact is it’s much more intelligent than viruses on computers that today we’re in awe of

03:52 of all the one. So, I do think that humans are going to have to make room for understanding but also harnessing machine intelligence to better understand nature’s intelligence and why not potentially vice versa. But does it mean that there could be a discovery that we we can’t explain at some point? >> There already are tons of discoveries that we can’t explain. It’s called the modern pharmaceutical industry. Probably more than half the half the drugs we don’t actually understand how they work. We can give one version of how they might work. But most of the drugs we don’t actually have any mechanistic

04:25 understanding of. But through rigorous testing, through controlled experiments, we could show that it’s more efficacious than it’s harmful. And so that’s regulated and it’s approved. So I I don’t think human understand by the way I don’t understand how my iPhone does most of the things it does doesn’t stop me from using it. I just don’t think this human conceit that anything we don’t understand we should be afraid of. That’s a new phenomenon. We’ve been coexisting with things we don’t understand for a very long time. >> Laya how do you see this? And actually if you look at deep mind I mean this is a lot of the work that you know you and Demis had started. Uh yes and actually

04:58 as as Nibbar was speaking I was thinking about AlphaFold as an example of a 50-year grand challenge and our advanced AI system uh Alphafold now we have it available to millions of researchers worldwide as simple as a map search to be able to look up and understand uh how the prediction of a protein a 3D uh structure which is remarkable because it tells us the has a basing build a basic building block of life and if it

05:29 misfolds what that might mean and we’ve been able to work with researchers like in Spain and taking one medication uh and applying it to the shagas disease an FDA approved medication uh impacts 8 million people worldwide something that as a human we hadn’t thought of but to ask the questions that helped us make that correlation between vaccines so I think the opportunity here is not only in the discovery of new things but also how we can apply other ones in fact there’s 180,000 researchers right here in India

06:00 working on problems that India is facing and advancing uh other types of work in understanding diseases and the medications that go alongside them. >> What are you seeing Kuran? I mean your what does it mean for Bioon? So are you expecting to how how does it change your company in the next five years? Yeah. So, of course, um I think drug innovation is going to accelerate, but I also hope that affordability or delivering affordable

06:30 access will also accelerate because you are shortening uh the timelines of uh you know lab to market. Having said that, I think whilst we are all focusing right now on the discovery process which is certainly going to accelerate and get abbreviated, the next challenge is how do you abbreviate the more expensive part of drug innovation which is clinical trials. And that’s where I again think that AI can play a big role in helping us think through what do we want to glean and gather from clinical data that allows us to you know predict

07:07 that it is safe and efficacious and I think those are the kind of things that can really compress timelines and deliver these patients faster to I mean um medicines faster to patients. So to me that’s the excitement of AI and that’s what we are looking at as a company saying how do we look at you know AI to really not just abbreviate but to also look at the affordability part for countries like India >> and I do want to come back to to the speed of it but actually to the the affordability do you think AI will make

07:39 the medicine just cheaper >> no as I mentioned I think it’s a it’s it’s a question of time is money and if If you can compress that time, there is of course going to be a cost element to a the use of AI itself, but hopefully the overall benefit will be that you can get medicines to patients cheaper and faster. That’s that’s my >> Yeah, no, I mean to that point, you know, you can have a discovery that a drug discovery that’s faster, but when do again, you know, the treatment of the patient gets faster?

08:13 >> Yeah. I mean I think there’s a lot of components to that that AI is helping and will help a lot. I think the the hardest thing will be to do what uh is out there this these days in terms of speculation which is simulating clinical trials and avoiding them altogether. I think that requires an understanding of human biology and physiology that I’m afraid we’re we’re quite far from. um albeit you know I think that we’ve certainly unders we’ve certainly modeled digitally lots of complex systems but I think most people who don’t work in

08:44 biology and medicine don’t understand just how complex these systems are. I assume our colleagues at DeepMind who worked on physics problems and biology problems learned what Schroinger learned 80 years ago which is that biology is not just physics and as you go up in scale the complexity of the problem goes exponentially high and in fact the folding problem is probably among the most kind of deterministic parts of biology everything else through interaction effects gets very complex so long way of saying I think there’s going

09:14 to be shorter term mto so for example which patient do you want to include in a trial? Can you train through their electronic medical records and find a subset that are going to be more responsive not just based on diagnostics but just based on their molecular profiles uh in cancer. I’ll give you another example. There’s a lot of interest in screening to find early cancer. The problem is those tests create 10 times more false positives than they do true positives. What are you going to do with those people? If you put them into the diagnostic system, it’ll drive cost through the roof. Can

09:47 you use AI to better triage and massively reduce the number of false positives? Those are things that I think as we start taking care of, it’ll start impacting clinical care and over time we could imagine more shortened trials, clearer endpoints. These are all things we’re working on. But I don’t think this notion of a digital twin of a human that we use for testing is something that that we should be looking for too soon. But what you’re talking about, what kind of timeline are we talking about? Like four, five years that the the

10:20 bottlenecks kind of unwind. >> I think the impacts being felt now. Uh and and it’s kind of hard to say in what sequence, but what I was talking about is this massive disruption in dramatically shortening clinical trials. I wish I’ve been doing this for 39 years. There’s nothing that would address the cost of new medicines as much as the clinical regulatory complexity and time. But I think that may that may be kind of a a step too far to represent to society and then disappoint people on. I mean, there’s

10:52 already enough with AI that’s causing disappointment and fear that I think going out and saying we’re going to cure cancer as some of the frontier lab people are saying as though it’s a proximal goal. I it’s really difficult to put that out there and then realize that five years from now you’re still curing. I mean, we’ve been curing cancer for 70 years. The joke in the biotech industry is that if mice had a cancer problem, we would have solved it completely because we’ve cured every which type of mouse cancer. Humans turn out to be dramatically different than than mice in their physiology. So we

11:24 have to be more humble. Let me say one other thing. I would say what AI has taught us in the last few years, the new versions of AI, generative AI, is that we now know there’s more we don’t know than we thought there was. In other words, as a fraction of what there is to be known, we now have gone backwards because we now know that it’s accessible. If only we could make measurements. if only we so I think that the fight between the unknown and the known is going to stay interesting for a while and that means I think as humans we have to be pretty humble in the way we go about doing it

11:57 >> how do you look at this yes it deserves some applause why not [applause] >> I’ll go back to what I said at the beginning I think it gets back to us just asking better questions asking questions that maybe we didn’t have access um to before as Kieran mentioned like we can take now an interdisciplinary approach like in the past where maybe we went really deep of being able to like look at patterns that may have been inaccessible. Um one of the things I’ve actually uh you know I’ve been in India for the past week and one of the areas I’m been really excited

12:31 to learn about um how we’re we’ve been collaborating here locally and some of the local innovation is around the area of agriculture. Uh so if I can shift maybe from the human side uh uh for a moment we’re working with um an organization called Icraat that’s based in Hyderrobad and one of the biggest challenges that we have is that sorghum which 500 million people rely on as a food source or to feed livestock uh has only really been studied in the west the western varietals not the indigenous ones here. So one of the challenges that

13:06 being faced in India is like how do we actually accelerate the understanding around the crop genetics and how to breed it. So we’re now working with crop breeders to look at sorghum or javar which I have learned my my uh word this week um to accel to to be able to map all the genes and make that freely available so the innovation can happen on site of that. Now, as mentioned by my colleagues, there’s a lot of work that needs to happen that follows that. But if we can have the breakthrough and get the inputs from the crop breeders, the

13:39 inputs from the farmers up front and accelerate, take a decade’s worth worth of time and start to crunch that together. Like imagine what this could mean for a more droughtresistant crop, something that can handle better diseases. And this is actually quite personal to me because uh my mom actually did sorghum research uh in Africa over 20 years ago and they got the world food prize for one of the first breakthroughs in droughtresistant uh sorghum but Africa and India are still facing so many challenges and it’s

14:11 really amazing to see now this what might be possible with some of the opportunities of applying AI into areas that have traditionally been uh underrepresented and underststudied. And I imagine the speed at the speed of adoption in that space is going to be lightning years faster than anything in >> Yeah, I mean we’re we’re hopeful and and in fact what we’re doing is that’s just on the seed side. We’re already working with farmers in a platform an AI powered platform called anthrochi uh with 600,000 farmers right here in India

14:44 about providing AI in based insights into the farmers that are relevant for the plot of land that they have. And what’s been amazing about that is through the 600,000 together we’ve worked in a very collaborative way including with state governments on trying to increase the accuracy and performance of this and uh in an open platform and have already exported it to nearly a dozen countries across Asia including Japan and Vietnam. >> Um Kieran could AI actually make personalized medicine accessible to

15:15 everyone? >> Well, it can. I mean if you look at uh the basic opportunity that we have of combining what I call biological intelligence and AI or artificial intelligence is very exciting. You know if you look at AI or machine learning it’s really based on uh you know logic and reasoning and pattern recognition on existing data or known data. Now what becomes very exciting is to Nubar’s point how do you get AI to actually do

15:47 something beyond that can it create new knowledge can it create new medicines and there I think understanding biology is going to play very important a very important role you know building bio biology foundation foundation models like what actually Nubar is doing at at Laya Science where you know they’re trying to combine uh machine learning with biology, biological data experimentation and that’s what biology does. Biology learns, adapts and then

16:18 inferences. I think inferencing is going to be very important comp compared to just reasoning and logic that is the continual realtime experimentation that you need to do learning these biology models and that I think will do a lot in terms of discovering new medicines [snorts] finding new you know sort of uh solutions to some of the age-old problems like neuro neurodeenerative diseases we haven’t fundamentally moved the needle when it comes through Alzheimer’s and and Parkinson’s because

16:51 we are trying to find a single point in the pathway the signaling pathway or a single point in in in the sort of the neural circuit or an immune kind of a network and that’s not working. So can we actually use these new systems to integrate and find very new therapeutics >> that can actually start making a big difference. So to me there’s a lot to be done but to everyone’s point we don’t know enough we’re just starting right and I think that’s the excitement that I

17:25 see in terms of combining all these various technologies >> is there a big question over trust so how do we trust the discoveries and especially how does the how do we make sure that the scientific community trusts the new discoveries by a >> I think trust is about the outcome you know if if you’re going to make a difference to healthare if you’re going to cure at least come towards kind curing diseases on maybe the word curing is not right but to make a big impact on diseases and disease outcomes then people will build up the trust and you’re talking about personalized

17:58 medicine for sure I think this whole database approach of getting the the genomic understanding of every individual and then personalizing the medicine to Nubar’s point of saying how do you make sure that you’re picking the right patients who will respond to that new medicine that is going to be your personalized uh medicine which I think is going to be very easily enabled by AI. >> Yeah. Maybe I mean I think it’s you’re raising a a provocative point because there’s trust in AI but there’s also

18:30 trust in science and at least in the US we’re living at a time where mistrust in science is a burgeoning industry and in science that what you’re trusting actually is is not just the results but the scientific method. Yeah. In other words, the way you get the result is hypothesis generation, falsification, and if you fail to disprove something, it’s considered the den truth. And then over time, if somebody else comes along and actually does an experiment that falsifies your hypothesis, you should celebrate because that means we learn something yet new. That methodology is what’s gotten us the vast majority of

19:02 what we understand to be today facts. Today we’re in an environment where we’ve gone backwards several hundred years where we’re basically debating facts with opinion. So people could say I don’t really believe in this and then they have you have to debate and they don’t have to present any facts. >> That erodess trust very very quickly in the scientific method. Let alone when AI comes in and adds to that where now it’s a machine that’s asserting a fact done the same way by falsifying experiments as opposed to another human. So I don’t

19:34 know. We’ll see if humans are going to trust machines more than they’re going to trust humans. But I can tell you we’re living at a time when somewhat postcoid in this last few years there’s been this real worrisome withdrawal from trusting facts. I I have a lot of doctor friends. I have a lot of pharmaceutical executive colleagues that are just happy for example they don’t work on vaccines because they think this is a problem limited to vaccines. This is not going to stop with vaccines. We will mistrust medicines. We will mistrust foods. we will mistrust algorithms. All we have to go on is is a methodology that as a

20:07 community humans can put forward like we trust the police because we’re living in a society where we’ve made a deal and some people might not but if everybody doesn’t you don’t have a society. I think it’s about as fundamental as that. So I think trust is almost now become a political concept and when it applies to science polit you know political science should not be the science kind of politics entering science and and as a scientist I’m very worried about that. >> How do you fix that because not everyone I mean I’ve you know some papers on science paper research has come through my desk I can’t read them. So a lot of

20:40 people can’t themselves know if it’s true or not. Well, generally I think that we have had people who are uh specialists and experts the collective view of which which will be easier to collect now with AI because there’s a lot more integration you could do represents the best you’re going to do when you go see your doctor and your doctor says based on everything I’ve know here’s what I think you should do and you go I don’t really trust you then where do you go next you go to another doctor but eventually you have to make a choice as to what what it is you trust so I don’t know there’s much we can do I

21:11 do think one thing and I’ve given given lots of suggestions to my scientist colleagues. I think scientists generally don’t enter the fray of actually explaining these things and actually scientists also have the characteristic that they doubt everybody else’s science. So we openly go to scientific conferences and challenge whatever people are saying. When somebody observes that they go wait a minute if this is challengeable then why should I trust it? They don’t understand that that’s part of the cooking process that results in scientific fact. So, it’s a complex area, but I do think we’re going

21:42 to have to as a society revisit where does trust come from, let alone unleashing AI, you know, move 37, all that stuff’s going to happen. And as humans, if we just distrust all of that, I don’t know where we go for our new medicines or our new a products or any of this. >> And we need to, I think, simplify science. You know, the articulation of science needs to be simplified to society in in general. And in fact that’s what we’re trying to do in a city like Bangalore where we’ve created the science gallery to basically do this connect society with science in a very

22:16 meaningful simple way that they can understand because to Nubar’s point I think that’s where it starts from people are confused >> but I also wonder like people trust the communities that they’re a part of and so much of this is about how do we start bridging between the real people and real problems and help them connect back into it. I mean, one of our approaches is very much of like you can’t develop the uh a platform for the farmers unless the farmer’s voices and needs are like built

22:48 in from the start. It can’t be an afterthought. Much like uh the work in the screening, you can’t think about how we’re going to do screening unless you have the doctor’s inputs in the beginning, not when we’re about ready to launch this. I think there’s something here which we really, you know, this is part of the reason why I moved into this role as uh chief AI readiness officer after years of building AI and the technology of and decades of working in technology as an engineer because I feel like AI is nent enough for us to shape

23:18 the future. It’s it’s made enough progress that we can actually see some of the real impact. But we have to work together if we’re going to co-imagine and co-create and shape where we want this to go, where we want it to be trustworthy, where we want to rely on AI, and where we actually want to focus on our human skills. I always ask everyone when you’re using AI, regardless of your whether you’re using it as a scientific tool or you’re using the generative AI in your home life, to really take a step back and think about

23:49 how you’re using it, but also where you and your bringing in your human ingenuity, your human creativity, your curiosity, your problem solving into that because I think those are the skills that we really need to be fostering. And it’s not just about what’s happening today like how are we preparing the scientists of tomorrow to ask the tough questions the ones that we might be limited in asking because we bring so much bias from our own lived experience both the wisdom of our lived experience but also maybe sometimes the limitations. >> Kieran I mean we talked a little bit

24:20 about you know the I guess the biotech industry in India. What are you most excited for your company? Well, I’m certainly very excited to leverage uh what AI has to offer and AI for biology, AI for science is something that we want to double down on because that’s a new area. I mean, I think we are just scratching at the surface, but it should give you deeper insights and you know the ability to connect dots which you can’t do today. So, I’m excited about

24:50 that. I’ve also I am excited about the fact that we have a pretty interesting talent pool in in a city like Bangalore which has a lot of AI talent and quantum biology talent and I’m just thinking whether we can actually combine that and do something more meaningful. To me the next opportunity for a company like Bioon is true drug innovation but leaprogging using these new technologies. Nubar, a question for you on first of all, would you regulate AI in science and if so, how? And then the second is

25:24 basically if scientific discovery becomes faster and cheaper, how do you build a successful biotech company? >> Wow. Two two separate >> two big questions in 20 seconds each. Um let let me let me address the just preliminarily the question about regulating AI in general because this is obviously out there and we have a particular exposure to it which is unique in that we live in the natural kind of living world as opposed to the digital inanimate world. And so we observe what’s happening in that world

25:56 from the point of view of understanding ecosystems and biology. And I could tell you that up until 2026 I was not worried about much of any of this. But in the beginning of 2026 with the early things that open claw mult books and some of these people showed which people quickly dismissed as insignificant only six months later despite a lot of warnings that some of us made that this is going to start looking like nature and it’s going to essentially start spreading agents that would create communities that would do unpredictable things. That’s precisely what’s happened. So I think that we’ve now crossed over into

26:28 not just large language models but with agents that that a large language model can allow you to generate with intentionality and then the ability to adapt so that it can now be misapplied itself not directed by a human in every which way that starts looking like a living organism that has adapting capability, swarming capability, ecosystem building capability. And people in the room, if they haven’t been worried so far, and I’m not a doomsayer for 50 40 years into this career, that’s not what we do. We’re optimists

26:59 technologically, but I think this is crossing a threshold. Therefore, I believe what we should do is to think carefully about regulating agents that are being commercialized and that are being transmitted outside the boundaries of what somebody may want to do in house. And we saw already with the hugging face example that these things can escape as well. But beyond that, beyond that, I think that if you’re going to do commerce of the agents, it should be incumbent on the people who are developing the agents, not the LLMs. To me, LLM’s are capability. Agents are intentionality.

27:31 >> If we don’t regulate intentionality, I don’t understand why we should be surprised if that intention of applies to us. And I think this can be done. this notion that you can’t regulate it or that if you do in the US the rest of the world to me all of that applies the same way to drugs as it does to AI agents meaning people didn’t regulate drugs very well until theide and in theomide people said holy cow and they all of a sudden with birth defects said they started overregulating and they said you know what the people who want to make money on this have to tell us why we should deem it safe if I was in

28:04 the AI space which we are in many ways I would start thinking about how to regulate and advance the reg sorry how to self-certify safety if you want to do commerce if somebody does it for their own use that may be harder to regulate but if you’re going to make money on it I think we have to think about it >> and there are also unintended consequences you know which require regulation because you will come up with new biioaterials synthetic biology is another area which we really need to look at very carefully so I think it’s both intended uh you know uses of bad uses of uh AI

28:42 and then the unintended >> LA it’ll be interesting to hear obviously you’re a bit in in a different position but it’ be interesting to hear your view on agents maybe and regulation of agents >> well you know one of the we’ve been long proponents of uh AI is too important not to regulate and and not to regulate well so we have to think if this is one of the most transformational technologies of our time how do we work uh to make sure that it is safe for humanity. Um, and you know, we’ve taken a number of

29:14 steps throughout the uh the decade plus of work that DeepMind has been doing around AI and working with policy makers to understand the technology itself. We’ve done a lot of work um specifically with around agentics uh agentic work workflows as well with in terms of ethics research trying to provide the community an opportunity to talk about the risks and the opportunities together. Uh so when we first demonstrated the technology alongside that we published our ethics research paper on the um ethics of advanced AI

29:47 assistance and I think Demis published something earlier this year in June about how we might go about thinking about how to regulate such technology and K just very finely because we’re out of time but you know how do we ensure today that some of these AIdriven breakthroughs translate into affordable drugs for patients is there is there a step that you need to take today? >> No, I don’t think so. I mean, it’s too early to, you know, I mean, yes, you can actually think about how to design it, >> but I think the main emphasis today is

30:20 really about how best to use AI to make a difference to what we’re doing today. >> Well, thank you so much for a great panel, Nubar, Laya, and Kieran. Thank you. Thank you. [applause] Thank you. [music]