Open AI Models, Agent Readiness and the Claude Conduct Debate

So it's been an interesting week as it relates to AI models, because there were two big pushes from what are largely being called Western companies that are creating open weight.

some open source models and the conversation is starting to change and Jeff I'm going to love your opinion on this but it seems like a lot of folks are now saying hey you know for all these awesome capabilities that these frontier AI models may have good enough is just fine and for 90% of the things that we need an AI model to do in the average corporate world well a slightly free or inexpensive model is going to be sufficient.

And now folks are saying, hey, we're going to see IT teams deploy multiple AI models and use orchestration tools to do that.

There's an article about CIQ's ability to do that.

But Mistral, and I think the other one is, is it reflection? Reflections beam, yeah.

And these things are, you know, at least from a conversational point of view, they're gaining traction.

I don't know about adoption just yet, but let me ask you something.

From your perspective, have we turned a corner here? Because I think a lot of folks are starting to realize, well, AI models, you know, the magic is not as complex as everybody initially may have thought.

So I do think we've turned a corner in a couple different ways, but not necessarily towards the, oh, hey, this shouldn't be as expensive or complex as we originally thought.

Because like anything, when you take away some of the rules, or I'll use the cliche guardrails, then you run the risk of what's it going to do.

And I think we need to recognize open way is not always the same as open source.

There are two pretty different things, actually.

And with a traditionally full-weight model, you have a more built-in bias, which may or may not be good, but you also have the built-in history of here's what...

History tells it, here's what makes sense.

Open weight, the good thing about an open weight AI is that it should be more free to be able to say, based on the data available to me now, this should be the answer, regardless of what we thought happened historically.

The downside of that is you may be treading new ground that may not be the right place.

And I think we run the risk of making it feel easy, which means that we can let our guard down about what we think the agents coming out of it are going to do.

And let's see how that turns out.

All right.

Jack, Paula, we were talking about the tensions around tech between the U.S. and China.

And part of the reason I think so many people are touting these particular models is they're not from China and people are less concerned that maybe they'll be trained to only respond to certain questions in a certain way.

Does that argument hold water with you? Is this something that the West is awake, as we say, with open source and open weight AI models? Or is this

just marketing? Well, it's not marketing nonsense.

It's driven by something else.

It's really not driven.