Eve Bodnia

And in terms of compute, you have to wait for like few minutes before it knows is it going left or right.

So you need to be able like on the seconds, milliseconds to understand what's around you.

I know, I know.

It's like maybe we are an AGI we're looking for and we're just trying to create a version of ourself, but...

Well, we're going to find out.

I can only speculate how it's going to be.

Right now, the scaling we're doing, it's all still on GPU for now.

But we haven't forced the model to be out there in the real world enough yet to understand

So for me, when I was working on it from the theoretical perspective, the most important question is, is the architecture scalable?

And what parameters are out there?

So there are ways to break complexity in your architecture without having too many GPUs.

I'm going to give you an example by what I mean.

So LLMs, they have some level of performance and

Then once you reach a critical mass of GPUs, so you have a lot of compute and a lot of billions parameters, then you have some sort of phase transition.

And all of a sudden you're seeing a different behavior.

And this is where things really start working for the LLMs.

There needs to be enough GPUs to see this kind of complexity change behavior.

And for our case, we see the phase transitions when we work with the hybrid, when we have the EBM attached to the LLM.

There are regimes when the LLM dominating, but then the EBM starts dominating.

And we just, like, if the EBM regime is dominating, then you don't really need GPUs.