Industry Collaboration in Advancing Clinical Technology

ACT: Could you describe the importance of collaboration, specifically through data sharing, with regards to advancing the development of technology solutions?

Stephen Pyke: If the technology is as powerful as we all hear it is, and indeed, it's proving to be the fuel for that technology, is ultimately data. Now, ChatGPT and the other generative artificial intelligence (AI) models are out there. Of course, what were they trained on? They were trained on a huge and very diverse set of data; material on the internet of all sorts of different characteristics and flavors. And what they've been trained to do incredibly well is, converse with humans, so they know how to make sense of a question. They know how to answer a question in a way that can be understood by humans. What they're not particularly experts in, at least not in this sort of incarnation in which we're most often able to engage with them as sort of citizens or patients, what they're not particularly expert in is my disease or the characteristics of patients like me and the needs of patients like me, and the sorts of treatments that might be most effective, given the kind of symptomology that I'm experiencing.

So typically, what we're going to want to do, in terms of many of the applications that might be relevant is expose them to new, specialized, rich data sources, which will allow them to be further trained. And this is something we call fine tuning. Which takes us directly to your question because many people may say, “Well, why do we need more data? They've all been trained already. They're already very smart. What else is there to learn?” It's that sense of they've been exposed, exposed to a broad and somewhat shallow ocean of data. What they now need is a more narrow, specific, and very deep and rich ocean that's relevant to the particular application or particular problems that we want to address on any given occasion. Well, then the next question is where is that data? And who's got it? And how do we access it? A lot of it sits with public bodies, governments, healthcare providers, institutions, and in some cases, insurance companies, but it's spread across a number of owners if that's the right word. In some cases, access is permitted. And there are mechanisms for allowing that. In other cases, it's more problematic, more difficult. And sometimes that data just doesn't exist at all. But the one thing that does strike me to your point is, we're entering a world, if we weren't already in it, where I think we have to see data as part of the common good. You know, I think none of us owns enough data, no matter how big of an organization we may be, to really do justice to the opportunity that AI presents. And so I think there's a real onus on those who manage, take care of, and are in a position to make accessible patient data. And it's that principle that we're talking about here. To think about how best to enhance access for the greater good of patients across the globe, who need access, who need the treatments that are going to flow from the approaches that AI can enable, and that would be very difficult without it.

So I think that's the straight answer your question. I would use this opportunity to highlight a couple of areas where collaboration also important, though. One is, I think, around the opportunity that is presenting to really create a strong framework for AI done well, for responsible AI, for ethical AI. No one organization can do that alone; I think there needs to be a collaboration among all the participants in clinical development to agree on what those principles are and we're starting to see some of those outputs: medical institutions, medical societies are beginning to put together their statements. ACRO, which is the Association for Clinical Research Oganizations, which the company I work for belongs to, they've created a statement of principles. Our own company in turn, Parexel, has created its statement of principle. But I think that this is a space where sharing our perspectives, collaborating on what good looks like is incredibly important, incredibly powerful potentially, and will do much I think to alleviate the concerns in the public discourse, which is around this all-powerful, all-knowing, all-seeing, a little bit scary technology. So I think this is a really critical area where we're already seeing very nice evidence of collaboration. And I don't think you can leave this area without also talking about the patients themselves. And the importance of seeing them as collaborators. Their voice needs to be heard. And it needs to be a part of the debate around what is the role of AI in clinical development? How best do we ensure that patients are beneficiaries, as well as the other participants in this whole process? So yeah, I think collaboration would be absolutely sort of central in the way I'm thinking about AI for data, but for other aspects as well.