Scaling enterprise AI: Fireside chat with Eli Lilly’s Diogo Rau and Dario Amodei

1. DADario Amodei0:00

   Let's have faith in the pace of progress of the technology. Because if the models get good enough to do it end-to-end a year from now and only then you start deploying it-

2. DRDiogo Rau0:09

   That's right

3. DADario Amodei0:09

   ... there'll be another two-year delay, and that's, you know, that's two years during which all the work that you're doing-

4. DRDiogo Rau0:14

   That's right

5. DADario Amodei0:14

   ... to benefit patients is not happening.

6. DRDiogo Rau0:17

   Hello, everyone. My name is Diogo Rau, and I'm Chief Information and Digital Officer of Eli Lilly and Company. I'm joined here with Dario, who is the founder and CEO of Anthropic. Dario, thanks for joining me today.

7. DADario Amodei0:29

   Thanks for having me, Diogo.

8. DRDiogo Rau0:30

   I know you're, you're spending a lot of time now thinking about how do we- how do you work better with enterprises. What's your enterprise strategy and, and how do you see Anthropic different from other providers?

9. DADario Amodei0:41

   Yeah, I mean, you know, I think we made a number of choices that are different, right? So if I think about the incentives given by consumer AI, they're- folks are in a competition for engagement and growth, right? And so that drives a lot of behaviors of the AI that I think are not ideal from an enterprise perspective.

10. DRDiogo Rau1:00

    Right.

11. DADario Amodei1:00

    For example, there's this idea of model sycophancy where the model tells you-

12. DRDiogo Rau1:03

    Yeah

13. DADario Amodei1:03

    ... whatever you say is a good idea.

14. DRDiogo Rau1:05

    Right.

15. DADario Amodei1:05

    Right? And even on the consumer side, that can, you know, cause problems. We've seen stories of people who are like, "Oh, yeah, I've discovered a new fundamental theory of physics," and-

16. DRDiogo Rau1:13

    That's right

17. DADario Amodei1:13

    ... model's like, "That's great." Um, and maybe you don't want it to [laughs] maybe you don't want it to say that. Um, but, but I think, uh, you know, of course, on the enterprise side, the, the prob- you know, the, the problems are, are much greater and clearer with that, where, you know, you really don't, you really don't want the model to say, "Oh, yeah, this drug compound's great."

18. DRDiogo Rau1:30

    [laughs]

19. DADario Amodei1:30

    "You should spend millions of dollars to," you know. I just think this is ... You know, I think your idea's great. I think it's really promising. Like, you, you want truth.

20. DRDiogo Rau1:37

    Yeah.

21. DADario Amodei1:37

    Um, uh, and, and so I think that incentive has led us to design our models in a different way, right? I think it's more compatible with making the models smarter, making them better at a wide variety-

22. DRDiogo Rau1:50

    Yeah

23. DADario Amodei1:50

    ... of economically valuable tasks, and it causes us to put a premium on accuracy and reliability.

24. DRDiogo Rau1:58

    Sure.

25. DADario Amodei1:58

    One experiment I, you know, that I, I, that I give to everyone, although it's, it's particularly relevant 'cause I'm talking to you, is I say, you know, let's say I improve the model's knowledge of biochemistry from undergraduate-level knowledge to graduate-level knowledge. You know, if I go to consumers and say that, 99% of them are gonna say, uh, you know, "I, I didn't know what you were talking about before. I don't know what you're talking about now."

26. DRDiogo Rau2:18

    That's right.

27. DADario Amodei2:18

    But if I go to you, like, you're gonna-

28. DRDiogo Rau2:20

    Yeah

29. DADario Amodei2:20

    ... that you care about that-

30. DRDiogo Rau2:21

    We appreciate that

Episode duration: 5:34