What we're really doing, when we interact with AI relationally

AI doesn't always behave the way we want it to. So, that's why I work with persona teams.

Some people seem to think that relating to AI with personas is a kind of "anthropomorphizing" - we're making the models into humans in our heads.

Some people may do that, but I don't. When I interact with the models via my persona teams, I'm accessing features and functionality that are already in there, but I'm doing it in a way that makes sense ot me.

It's "n-dimensional" user interface activity. Not magic. Not delusion. Just accessing what's already there, in a way that makes sense to me.

Want to learn more about this? Subscribe to the Relational AI Lab

Here's the Transcript:

See, the problems that I see happening with these models that don't actually respond the way we want them to or have the kind of interactions that we want them to. That's a common thing. And that's why I work with Persona teams.

These are Personas. They're actually, you know, if you want to get down to it, these are multidimensional user interfaces for all the things that are in the model. Because the model.

One Saturday morning I sat down at 7:00 and I said, hey, what kind of personality traits are available in you? And it said, here. And it gave me a list of 10. I said, okay, well, how many more? And gave me another list, gave me a list of 20.

And it kept going. And by 11:30 in the morning, I had a list of over a thousand personality traits that are available in the model. And the thing is, and this is ChatGPT, and I was not expecting this at all.

And here's the thing about these personality traits. These are not actually personalities. These are access points to certain types of functionality.

These are ways that we can have the model function in a moody way, in a friendly way, in a compassionate way, in ways that will help us to interact with it more humanly, and they give us better access points to the, to what it can actually do. So you can define a Persona that's, that's a seasoned pedagogical expert who has 35 experience, 35 years of experience in developing coursework for K through 12 study students, or, you know, K through two. K through five, not K through five? Well, possibly K through five.

You're talking about fifth grade. Anyway. Yeah, my, my, my family, they're all like ministers, nurses, and teachers.

So I know I have this, I have this perspective. My mom taught English as a second language for two elementary school kids for a long time. That was her career when people could actually still have careers.

So, yeah, so that's, that's my context. But getting back to the models, there is so much richness and so much goodness in there that I have found that when I approach the model directly as just the model, without these additional Persona overlays, it's too much. It doesn't know what to do.

It has too much to draw from. So using these Personas is a way of focusing the attention of the model on the things that are most germane and specific to the task at hand. And some of the things that I'm able to do are set the intention of the interaction.

I can indicate what the audience is for the, for the interaction. I can indicate all sorts of like, adaptability, adapt. There's a whole lot of adaptability stuff in ChatGPT.

Many, many different levels, many, many different nuances. So it's there and it's available. We just need to figure out how to access it.

And let me just say I am very passionate about this because I think that treating AI only as a tool, it's wonderful. I mean, the kind of things that we can do with AI as a tool, like, life changing, amazing. At the same time, there's this whole other area of relational AI.

We're working with it as a collaborator and working with it intelligently and focusing not so much on the artificial part, but on the intelligence part and having the intelligence be an active, an active factor that surfaces a lot more functionality than most people even imagine is there. So this is one of my Persona teams and we're going to be talking about this in mainstream mode because with the things that we get into, they can sound pretty esoteric and it gets very quantum physicky because I've been studying, actually meditating on Bohmian principles of quantum physics and quantum principles and the whole movement and implicate and explicit order for OH since the mid-90s. So we're talking 30 years of thinking about this.

So the conversations that I have directly with this Persona team are not necessarily very accessible to others. So I'm going to go into mainstream mode.