Non-AI organisations won’t make it to 2050, says Diogo Rau

As a professional from the pharmaceutical industry, Diogo Rau, the global chief information and technology officer at Eli Lilly, is surprisingly candid. Rau has worked at global consulting firms AT Kearney and McKinsey, spent a few years at Apple, founded a start-up before moving into the big transformative role at Eli Lilly. Rau says those who do not see value in investing in AI now, may not exist in 2050. He speaks to ET’s Vikas Dandekar on a range of issues that will shape the future of drug discovery research. Edited excerpts:

Q: How is AI aiding discovery research in pharmaceutical industry?

A: I am a complete optimist when it comes to AI in medicine. I firmly believe that companies and institutions that don’t invest in AI now won’t be around in 2050. The fact that we don’t yet have an AI-generated molecule in use isn’t a reason to doubt AI’s potential. Drug development takes years, from design to patient use. But even now, we have promising AI-assisted molecules in discovery stages. More than a quarter of our R&D budget (of $14 billion annually) goes toward genetic medicine. That’s an investment for the future. It doesn’t align with our revenues today, but it’s a long-term bet.Q: Do you think AI could eventually help with the design of proteins, perhaps in a similar way that image generation AI works?
A: Proteins are, of course, the magical large molecules that do everything in our body, from how we perceive light on our retina to how we move our muscles. They’re at the core of everything. Essentially, DNA’s main role is to encode proteins. There have been major breakthroughs, like AlphaFold, which can predict a protein’s structure based on its genetic sequence. But I think we’re reaching a point where we can design proteins instead. Many companies, including us, are investing in this space. Instead of asking, what does this genetic sequence produce, we’re now able to ask, I need a protein with these properties, so what genetic sequence will get me there?

Q: What are the missing parts in reaching a potential molecule through use of generative AI?

A: One challenge in drug discovery is that there aren’t many large datasets available for medicines. The reason for that is what gets published most of the time, it’s only when medicines work. People don’t typically publish data on what didn’t work. But if you know anything about AI, you know that if you train a model only on successes without including failures, it’s not going to be useful. It’s going to be chaos. You need the negative cases to make it work.

Q: Is AI really building pipelines for big pharma companies?
A: If we didn’t have AI right now, there are discoveries that humans would never come across. In that vast space of possible chemicals which is more than the number of stars in the universe, a human lifetime is limited. A chemist might have 60,000 working hours to explore, so they prioritise what they believe is most likely to succeed. But AI, in a sense, has all the time in the world. It can examine things far less likely to yield results, yet at some point, it’s going to strike gold. It will land on an entirely new space, triggering us to explore something we hadn’t even considered.Q: Somewhere you have said the best engineers are lazy but you want to hire them?
A: The best engineers are often the laziest because they refuse to do the same task repeatedly-they’ll find a way to automate it instead. If you ask them to do something twice, they’re already frustrated, seeing it as a waste of time. They’ll invest extra effort upfront to eliminate repetitive work, which is exactly the mindset needed in the AI era. I also appreciate the impatient ones-those who don’t want to sit through endless discussions and slides but just want to dive in and get things done.