Logan Kilpatrick

So I think you'll need,

These sit like human memory has this like interesting has all these interesting mechanisms to make sure like the stuff that's not useful is removed from your memory and the stuff that is useful cemented and then you build these this sort of abstraction layer on top of.

um like core ideas that you've had i assume we'll get something closer to like what looks like memory um because then it it decreases like you don't need 100 million you probably need like 10 million with the system on top of it um and that's much more reasonable than 100 million totally that makes sense

Yeah.

This depends probably a little bit on like the use case.

You could definitely do something like that.

Um, I would, I would suggest folks to like, try to do some validation as far as like how accurate that is.

And I imagine there's cases where it works really well and there's probably like foods, like different food genres or categories or different examples where it does not work well.

Um, yeah.

So this is where making it easy for people to build evals and benchmarks is really important.

I'd love to see your personal eval of different food understanding use cases.

But yeah, I do think I've thought about this for...

a while um if you've ever like tried to use one of those apps to like track what you're eating um it's a huge pain in the ass i'm like i i envy people who are going on that journey and that mission because it's such a pain to have to like keep track of everything and being able to just snap a picture and have the model deeply understand um and use things like bounding boxes and it can do it the size proportional understanding of size and all the like nuanced details

um, has just never been possible without language models that have like a deep vision capability.

And I think that's, that's something that works like actually works right now.

And it is a good example of like the model capability is actually in some cases, like, uh, outpaced the ability for us to build products around the model and the ability of people to like wrap their head around the understanding of the model can actually do that.

Like if you were to go and ask, you know, a hundred million random Americans, can you

Does a product like that exist?

Can AI models do this, et cetera?

Like I would assume the vast majority would say it's not possible.