Ideas to Innovation - Season Three
Henry Levy: The potential of gen AI it’s almost like a ship taking us to a new galaxy. I think for the next three to five years, it’ll take us to new planets which is still amazing. What we’re going to see in the pharmaceutical industry, in the healthcare industry, over the next two to four years, I think will be considered transformational in that it will accelerate things, it will significantly create space for new activities to be done because of efficiency, it will transform the workforce pretty significantly.
Intro: Ideas to Innovation
Neville Hobson: Artificial intelligence is revolutionizing our world in many different ways. It’s not a single technology. AI encompasses an abundance of technologies, including machine learning, deep learning, and natural language processing. In life sciences and healthcare, these technologies are instrumental today in tasks such as diagnostics, precision medicine, and even robotic surgery. The transformative impact of AI is already being felt in areas like pharmaceutical research, epidemiology and telemedicine. There are projections that the global market for AI and life sciences will be US$6.7 billion by 2030. What’s driving such growth and how rosy is this picture?
Welcome to Ideas to Innovation, a podcast from Clarivate with information and insight from conversations that explore how innovation spurs incredible outcomes by passionate people in many areas of science, business, academia, technology, sport and more. I’m Neville Hobson.
Joining me at the start of this new season three of Ideas to Innovation is Henry Levy, our guest who has plenty to say about AI and what it will and won’t enable us to achieve. Welcome, Henry. Thank you for joining us.
Henry Levy: Thank you for having me Neville.
Neville Hobson: So you bring a distinct outlook to our conversation as your president of life sciences and healthcare at Clarivate. You joined Clarivate earlier this year. Let’s put your role in context for our conversation. What got you into this career and brought you to Clarivate? What’s your passion?
Henry Levy: Well, I think we’re gonna have to go way back Neville. My parents are both in the medical field. My father’s a neurosurgeon, my mother’s a nurse and I always wanted to be in this space. And for a long time I thought I was going to be a medical doctor. And in my head I thought, how can I help more people? As a medical doctor, you help the patient in front of you. But I always thought that we could do more if I could be in the health arena. And then after university, I decided that the pharmaceutical industry was one that I think could have the most significant impact. It was personal to me. I had some medical issues as I was going through college. And I found a drug that I’ve been taking for 31 years, and that has improved my quality of life immensely. I would not be able to have the life that I had if I didn’t have that one pill that I take once a day, and that has been transformative for me. And I would assume that our audience. also has many, many stories of a drug that has saved a parent’s life, a friend’s life, a husband’s or wife’s or partner’s. And therefore for me, being in the pharmaceutical and life sciences space was clear. So I went into it through the consulting world. I was at Accenture for 21 years. And then after consulting with the largest pharmaceutical companies, I decided to try to be in the services industry in that space, and I joined a CRO, PPD. Once I did consulting and services, I felt that technology was the most transformative part of our industry, and I went to Viva. And then about four months ago, I decided that the one thing that completed that story was data. The pharmaceutical industry is about data. It’s of course about creating drugs that help patients, but the way that you prove that… is through data and Clarivate is by far, I’ll say the most advanced company in bringing together technology and data to help the life sciences industry. So it was a natural fit for me.
Neville Hobson: Thanks, Henry. That’s a great scene setter. Let’s be clear on what we’re going to talk about in this conversation, which is generative AI. Gen.ai, as some call it, is a form of machine learning that’s able to produce text, video images, and other types of content. ChatGPT and Dali are examples of generative AI applications that produce text or images based on user-given prompts or dialogues. So it’s not just AI. I mentioned earlier it’s a collection of technologies. This is generative AI. I think it’s fair to say that the integration of generative AI in life sciences faces challenges. Regulatory hurdles are paramount, ensuring the safety and quality of AI driven products and safeguarding patient data privacy are areas of concern. And according to a new report from McKinsey published in August, key risks like inaccuracy, cybersecurity and IP infringement are not being sufficiently addressed yet by most organisations. McKinsey does note though, that generative AI adoption is happening quickly, a third of organisations already using it regularly in at least one business function they say marketing, product development and customer service are the top use cases. Now I know Clarivate has been using AI in its products and services for quite some time. Can you tell us how generative AI plays a role in life sciences and what challenges has Clarivate faced in integrating this new technology?
Henry Levy: It’s a really good question, Neville. I think that generative AI can and should be transformative in the pharmaceutical industry. But I think that transformation is going to be focused on the operational and efficiency side, and of course, driving speed in drug development and in many other parts of the pharmaceutical industry. But the more transformative side of the equation, where it might generate opportunities, generate drugs, generate candidates, is probably a five to ten years from now perspective. I think that today, if you think about the inputs and the outputs of artificial intelligence, they’re not validated in a way that you could really guarantee that the output is appropriate. And the regulators, if you think about them, have taken a long time usually to provide guidance as to how to use those components. Whenever you, let’s say, are submitting a drug, the authorities want the provenance of everything that you do. Provenance means the data, where the data comes from, where the results come from. And if you’re creating something through generative AI, it’s going to be difficult to prove where it came from. So we’ll need some guidance from the FDA and from the European authorities as to how you test and validate those models in order to be included in the submissions that you need. So from my perspective, the industry is going to be able to shorten the timeframe because you’re going to have generative AI creating regulatory documents, that it’s going to be drafting them, but you’re going to have a step at the end where a human will have to validate it, will have to check it. So what you’re going to see is a significant contraction of the timelines. which is going to be great for the industry, but the actual identification, the generation of intelligence associated with the drug, I think will take a little bit longer.
Neville Hobson: That’s interesting you mention that because I suppose you could simplify that in a sense by saying that this will take the kind of drudgery out of labor intensity in research, for instance, where you’ve got a tool, a technology tool that has the ability to process structured, unstructured data, whatever type of data, rapidly at such speed. Humans can’t match that. And so that means we would have faster… go to market possibilities you think?
Henry Levy: Well, first of all, being in this industry for 30 years, I would prefer not to say drudgery. And the reason I say that is that work is important. That work that is happening today, being done by humans, experts, actually, in many ways, is very valuable. And it’s more a really a call out to the amazing technology that we’re now introducing, that it can actually do a lot of the things that we used to do as humans.almost as good as we can. It’s a good thing, but it’s not drudgery. This is not about just executing a repeat, a repetitive activity. These are pretty difficult tasks, but the technology is actually caught up with that. And it will definitely, I think, shorten the timeframes. The main issue from bringing drugs to market is that in the clinical world, you still have to treat a patient. So if you think about the timeline for bringing a drug to market, which can be up to 10 years, there’s only somewhere around three years that can be cut off, as in activities that are being done outside of the clinical trial process. So I think those edges are going to be compressed, but artificial intelligence is not going to do anything when you have to give a drug to a patient and wait for that patient to be treated. Let’s say it’s an oncology trial. You need to see if a year or two or three go by and there’s not a recurrence of the cancer. And therefore that timeline is not gonna be compressed. So I think there will be some acceleration, but you still have to do the clinical trials and AI is not going to affect that.
Neville Hobson: Got it. Okay. Not drudgery. I note that very carefully. It is not drudgery.
Henry Levy: I’m just defending my cohorts.
Neville Hobson: Yeah, that’s good. That’s good. So just going back a bit to what we talked about at the very start where we were saying that Clarivate’s been doing this for quite some time. Can you share some specifics on what Clarivate has been doing and some idea of what customers have been able to do as a result of this work?
Henry Levy: Definitely. As you mentioned in your introductory comments, Clarivate has been using AI for six or seven years. In all of our products, there is some level of machine learning or artificial intelligence that looks at past data and makes a prediction or it makes a decision going forward. But just in the last two months, we’ve introduced new capabilities that are generative in nature. We are a data company. One of the ways we use that data is for competitive intelligence and market intelligence to help our customers make decisions as to where to go, which markets, which diseases, which products to decide on. And what we have done is we have thousands of reports and massive databases that today, or let’s say in the past, you had to access, you know, through searches. Now we have the ability to do what is called a question. Our customers can ask a question, and the model can look at all of that data across structured and unstructured data and come back to them with a clear answer to their question, but also point them in the direction of the reports and of that data that they need. So it will accelerate the actual result, the outcome, but also point them in the direction of where that data came from, which is really critical for their decision making. So we’re really excited about it. Our initial customers are saying that this is transformative, but it is just the beginning. From our perspective, if we can leverage data in a more effective way, we have data across epidemiology, which you mentioned, toxicology. We have historical data about the pathways that the chemicals take through the body those can be inputs into target identification. We may be able to point not just which products are good from a market perspective, but which ones we can predict will perform better based on past performance. So we’re very excited about where we’re going.
Neville Hobson: Let’s talk a bit more about some of the challenges and hurdles in this marketplace. You mentioned provenance of data earlier. And it makes me think of something else I mentioned prior to that, which was safeguarding patient data privacy, that many people say this is a major area of concern in this embryonic industry. Is it, I wonder? Maybe as a kind of general term, but specifics account a lot here. So how do you see it? Is this something that anyone should be seriously concerned about?
Henry Levy: The way I would answer is that every single company that’s playing in the space needs to be seriously focused on it. I don’t think that the patient population or humans in general should be worried about this. I think that there are government protections, there are industry protections, there are certifications. And I truly do believe that. privacy is if everybody is working effectively and in a legal way, privacy will be maintained. But it requires focus, it requires attention, it requires investment, and therefore you have to go through all of those components. The part that I worry, there’s two parts that I worry more about. One is you mentioned intellectual property. I think that still understanding how the data… the assets that get introduced into our large language model are compensated for, are used for, is still to be figured out. If you think about it, we, on the Clarivate side, we license our data. Our customers pay us to license that data. Well, if that data is used in our large language model, you know that it will learn from it. What happens when the customer stops paying for that license? Do we need to remove that data from the large language model, which is actually unlearning something? So I think you’re going to see a generation of new models. The second thing that worries me, in the pharmaceutical industry specifically, is that our models are learning from our successes and not from our failures. So by that, I mean that there’s a massive amount of data when a drug actually works. It is submitted to the authorities. There’s a lot of information that’s provided. But if you think about the, I’ll say the funnel of drugs going into the pharmaceutical industry, you may start with 100,000 candidates to get to one. And we need the failures in order to learn. But that failure information is not easily available. So the models that you would have today would clearly be optimistic, as in it would have most of the data about the things that succeeded. and it would exclude all of the data of the failures. So we need the industry to come together to start thinking about how does it share those failure data sets in order to make the large language model truly complete so that it doesn’t just learn from the successes but also from the failures. Does that make sense?
Neville Hobson: Yeah, it does. It actually leads into, I think, another element in all of that too. In this, what do we call it, this age of AI, I suppose is a good moniker to give it. Is this overall environment fit for purpose, such as we’re discussing? Maybe it’s not the environment. Is the AI fit for purpose, given the possibility for, what would you call it, not wrongdoing exactly, but the stories you hear, people are concerned about, hallucination for instance. But I’m thinking more of the automation element of it. And I’m wondering what your thoughts are on that, Henry, about AI being fit for purpose in the areas we’re discussing, i.e. healthcare and life sciences. What do you think?
Henry Levy: Well, I think that CHAT GPT and GEN.AI was a little bit like COVID was for the industry. COVID actually moved the industry forward. It accelerated it because it was a very, I’ll say, impactful experience. CHAT GPT has been that for the pharmaceutical industry, and you see a lot of companies trying to do things. But I think McKinsey, the McKinsey study that you quoted, highlights the fact that a relatively small percentage, about a third, are actually focused on solving the issues associated with it. So I do think we still have a lot to do to ensure that it is, in quotes, safe. And therefore, I think that when we get into true automation of activities, I’ll say equating it to a driverless car, there’s gonna be questions about who’s accountable when there’s an issue associated with an AI engine. Is the company accountable? Is this source technology accountable? And therefore, I think you’re going to see a new set of criteria, a new set of rules, a new set of behaviors and even legal precedent that will need to be defined. And because of that, I think you’re going to see that it’s not fit for purpose. I think as I mentioned in my first commentary, it is fit for purpose for cutting costs. for reducing the timeline, for allowing what you mentioned as potentially easy activities to be done faster. But I don’t think you’re going to see it as a highly transformational part of the industry until all of these things are figured out in our industry and in many, many others. I wanna be clear that gen AI is going to be applied today and it’s gonna be applied tomorrow and it is going to have a massive impact. So I think we discussed this at some point. The potential of gen AI is to, it’s almost like a ship taking us to a new galaxy. I think for the next three to five years, it’ll take us to new planet. which is still amazing. What we’re going to see in the pharmaceutical industry, in the healthcare industry, over the next two to four years, I think will be considered transformational in that it will accelerate things, it will significantly create space for new activities to be done because of efficiency, it will transform the workforce pretty significantly. But what I’m saying is that transformation down to, can it cure cancer? Can it solve the actual… underlying issues of drug development, I think those will take five to ten years. So we should be incredibly excited. It will be transformational. It’s just, you know, the largeness of its impact could take us to a new galaxy.
Neville Hobson: That’s actually a really good snapshot view looking ahead. Let me ask you one or two more things about that picture you’re projecting. So we know the landscape today, that’s really what we’ve been discussing, leading into what’s next, what’s coming next. And we have this, you know, 10 years out is not a bad period to be looking at. 50, most people can’t relate, that’s too far away, but 10 years, yeah, that’s 2033, that’s the next decade, which is approaching quite fast. You’ve mentioned some elements here. What would you want to say in the context of these changes you’re talking about? What’s it going to let companies do, particularly customers of Clarivate for instance? What amazing things are going to be available to them 10 years from now based on what we know now? What’s your thinking on that?
Henry Levy: So let me just hit it from two or three different points of view. I think from a regulatory perspective, if you think about it, if Gen.AI can generate, I’ll say, 99% of the reports that need to go to the authorities, I think that it’s going to create a revolution for the regulators. As in the regulators are going to say, why am I getting… automatically generated outcomes. I shouldn’t do that anymore. And I think it’s going to push the regulators to change the moniker, change the model. And instead of saying, you pharmaceutical companies create data, create reports and send them to us, it’s going to create collaboration. And you’re going to see the regulators actually working directly with the pharmaceutical industry and allowing for approvals in days instead of years. So I think it’s, it’s not that Gen. AI is going to accelerate, but the gen AI is going to create an environment where the authorities say, why am I waiting for a computer generated result when I can just collaborate with the industry directly? So I think that’s going to be an interesting dynamic. On the discovery side of things, I think that the number of products that we will be able to discard quickly and then hopefully pinpoint down to a smaller number that have to go through preclinical and clinical development is going to be massively accelerated. I think you’re going to have marketplaces of, I’ll say, AI engines that will start to generate opportunities for the industry to pick up and go. So you’re going to have more opportunities or more shots on goal. And then lastly, on the commercial side, I think you’re going to have the… patient population and the physician population being able to get access to information in a more direct way and therefore creating a more informed population, a more self-directed population from a physician perspective, which will actually accelerate the use of drugs in a more effective way. So I think you’re going to have an impact on discovery, regulatory, and the actual use of the drugs themselves.
Neville Hobson: Exciting times ahead then, I would say. That’s really, really fascinating. This has been a great conversation, Henry, in the time we’ve had available to do this. And we’ve only just scratched the surface I have in my mind, frankly. There’s lots more we could have talked about, maybe another time perhaps.
Henry Levy: Can’t wait.
Neville Hobson: But thank you so much. Thanks for sharing your time and your insights. This has really been great.
Henry Levy: Thank you, Neville.
Neville Hobson: You’ve been listening to a conversation about the impact of generative artificial intelligence on the life sciences and healthcare industries with our guest, Henry Levy, president of life sciences and healthcare at Clarivate. For information about life sciences at Clarivate, visit clarivate.com and choose life sciences and healthcare from the top navigation menu. To find out more about AI in Clarivate, visit clarivate.com/AI. We’ll be releasing our next episode in a few weeks time. Visit clarivate.com/podcasts for information about ideas to innovation. And for this episode, please consider sharing it with your friends and colleagues, rating us on your favorite podcast app or leaving a review. Until next time, thanks for listening.
Outro: Ideas to innovation from Clarivate.