Arthur Mensch

And so at the beginning, synthetic data allows you to do the compression, to further compress the models.

And so at the beginning, synthetic data allows you to do the compression, to further compress the models.

At the end, you do need to go and get data that is produced by humans.

At the end, you do need to go and get data that is produced by humans.

So yeah, it's a way to have... It's mostly an efficient way of training models to have bigger models that are used as teachers for smaller models, but it's not enough.

So yeah, it's a way to have... It's mostly an efficient way of training models to have bigger models that are used as teachers for smaller models, but it's not enough.

And so you also need human signal.

And so you also need human signal.

What does that OpenClaw moment mean?

What does that OpenClaw moment mean?

Well, it resonated a lot with what we were doing with our customers because pretty quickly enterprises realized that if they wanted to make some gains with artificial intelligence, generative AI, they would need to automate full processes.

Well, it resonated a lot with what we were doing with our customers because pretty quickly enterprises realized that if they wanted to make some gains with artificial intelligence, generative AI, they would need to automate full processes.

And to automate the full process as an enterprise, well, you can use OpenClaw, but it's going to be, it's actually not really enough because

And to automate the full process as an enterprise, well, you can use OpenClaw, but it's going to be, it's actually not really enough because

You have data problems, you have governance problems, you can't observe the process that is running, and you can't control it.

You have data problems, you have governance problems, you can't observe the process that is running, and you can't control it.

In many cases, when you run a KYC process, if you're HSBC, for instance, one of our customers, you will want to have deterministic gates that are going to always do the same thing in a way that is observable and that you can guarantee the CAO that it's always going to go through these gates.

In many cases, when you run a KYC process, if you're HSBC, for instance, one of our customers, you will want to have deterministic gates that are going to always do the same thing in a way that is observable and that you can guarantee the CAO that it's always going to go through these gates.

And that's not something that OpenFlow is providing because it doesn't have the kind of primitives that you need

And that's not something that OpenFlow is providing because it doesn't have the kind of primitives that you need