Julia Shaw

And so you get confabulations from both sides now.

You get confabulations from AI and from the people.

And the problem is that there's a third thing, which is the in-between that I'm not sure is getting enough attention right now.

And I wish that there was more integration of social scientists like me and people who do investigative interviewing and have done it for decades.

to understand what is happening in the in-between, and so that we can both teach the people and the AI to respond better in that situation.

Yes.

There is...

What we've created with GenAI is basically the ultimate false memory machine.

We have created a tailored experience of something that is most of the time telling you what it thinks you want to hear

And then it's uncritically giving that to you.

Or I mean, sometimes, of course, there's, you know, other things where it's sort of appraising whether or not this is truthful or not.

But it is giving that to you.

And there's no safeguard from you just going, this is truth.

And this is my past or this is how I remembered it.

And the problem is, is that not only is AI potentially distorting people and their memories, and never mind the factual basis on which they're relying, but it's also the other way around, is that potentially by asking leading or problematic questions, the people are changing how the AI is creating the content, which is in turn on some fundamental level potentially having an impact on how it's discerning truth from fiction.

And so that's where the false memory in human minds and confabulations in AI, I think are much more similar than we think.

And when I first saw AI confabulate, hallucinate, I was like, this is what people do all the time.

It's just that we can't fact check them all the time.

We're not in a conversation constantly being like, well, is that quite right?

I'm going to use that for my homework, right?