Antoine Papiernik

To be able to be successful in tech bio using AI, you need to have an understanding of healthcare and have models that ultimately are dedicated to this space, leveraging the data in our space.

And I will argue that there's no space

like drug discovery, drug development, where data is more important than, you know, than other industries.

If you think about LLMs that are based on, you know, the 26 letters of the alphabet, think about the complexity of biology from, you know, DNA to RNA to proteins to ultimately, you know, all the connections between cells.

This is just data like, you know, it's paradise for AI, but it needs to be developed specifically.

I'll tell you, for instance, we have two companies in our portfolio that are the next generation of foundational model for biology.

I can mention one in France, which is called Bioptimus, that is trying to decipher, to do the same as the LLMs, but taking biology as the building blocks.

And then we have a company in the U.K.,

called Latent Labs that just announced, I think yesterday, their Latent X2 model, which really looks like, could you, using AI, develop tomorrow's next generation proteins and develop antibodies that for sure don't have immunogenicity from day one?

If you think that leveraging the knowledge of AI, but in using the data in our own sector,

And this will require, I mean, ChatGPT, I mean, OpenAI and the others, and Tropic can do a lot because they have hundreds of billions at their disposal, but they lack the knowledge of our industry.

Of course, they could buy, but I predict that, in fact, they will ultimately buy potentially or partner with those who have developed over the next three, four, five years, these technologies dedicated to our field.

Let me just say one and take another cut at this, which is our own industry, the venture industry, the life science venture industry.

It happens that six years ago, six or seven years ago, pre-ChatGPT, I had no idea what was going on there.

We decided to

become data savvy, maybe at last, and then use, you mentioned the fact that we were set up in 1972.

So use the fact that we have all these data over 50 years, annotated information about deals that we saw, that we studied, where we have consultants from the FDA, from reimbursement, from whatever, CNC, and then use that

and connect those dots with all the public dots that are available out there about mechanisms of actions and clinical trials about patent.

So we've done that.

That was six years ago.