Caspar Eliot

Pleasure to be here.

Thank you for having me.

So, Caspar, Invisible Technologies just raised $100 million.

Invisible also says it has trained 80% of the world's top AI models.

Both of those are incredible stats.

Can you just tell us a little bit more about what that process actually looks like?

It's what you're after, but then it's making stuff up.

When people use ChatGPT or Claude, they have no idea that there are humans in the background who have labeled millions of training images, rated response quality, as you were saying with human feedback, fine-tuned models through RLHF, and even reg-teamed them to find out biases.

So all of that goes into this process, I'm assuming, correct?

I saw that.

Yeah, that was recent.

So you mentioned a minute ago, you mentioned that you think that there's a lot of value in the application layer.

And as someone who's training these models, working on this, do you then see the future being more so around smaller, more niche models?

Do you see that being more of where we're going to get a lot of the performance gains in the models?

huge trillion dollar companies spending a lot of money on their base models because their base models are behind the application layer and they're going to be behind a load of other companies application layers let's pivot back to invisible because um i know that you work with a lot of companies and uh i want to know a little bit more about how it how it works when you work with a company uh for instance there's like the frontier labs but there's also like let's say um you know in your recent fundraising uh video you had an example with the charlotte hornets came up

And how would you work with a company like that with your process, if you could walk us through how that works?

So as part of that, what did the humans do in that process versus what's automated?

In human data or in?

Yeah.

Or I guess like you have quite a few different services that you offer, right?