Experiment

Mm-hmm.

And one of the key, one of the key behaviors we'll have to em- embrace is experimentation.

So we'll have to give people the permission to say, "You don't have to have all the answers," even though, like, some people have master's degrees and PhDs and, you know, all the school system is based on us having all the answers.

But now we're saying-

Sorry, Mary, it just links back to the, uh-

No [laughs]

... experimentation piece.

Sorry, I'm talking too much.

And some of the themes that I'm looking into is how do we establish that pipeline in a way that, you know, it is going to help people rather than hampering your process.

Is there a bit of a... an issue that where AI agents are being deployed and perhaps they're hallucinating, there's an element that there's a lot of experimentation going on, there's a lot of shadow AI about, and 90% of enterprise data is unstructured.

So you put an AI agent into that kind of an environment where Experimentation is happening beyond governance and the data is not ready.

Well, obviously it's going to start hallucinating.

So organizations aren't really laying the foundations first and then they're seeing the kind of results that they wouldn't want.

Some of the finest data we have on it emerges from the parapsychology lab at Duke University.

It was amassed during the early stages of the Second World War, where Experiments were conducted, thousands upon thousands of trials, vetted, juried, bulletproof, meta-analyzed.

In fact, it was that lab that minted the concept of meta-analysis, where you're grouping together different experiments, but similar, into one study and trying to analyze the possibility of chance or accident.

These experiments, these psychokinesis experiments at the Duke Parapsychology Lab demonstrated the ability of thought among certain subjects to influence the throw of dice.