Allie K. Miller

Knowing that there are ways to improve it.

We can ground it in information.

We can use more state-of-the-art models.

We can give it access to the internet, check citations, all this stuff.

Yep.

Why does it still BS us?

Why does it still hallucinate?

You just hit the nail on the head, which is we told these systems, be helpful to me.

And these systems converted that task and said, oh, you want me to be helpful?

You want me to always answer because when I say, I don't know, that's not helpful to you.

So because these systems were not given an off-ramp to say, you know,

They're not allowed to say, I don't know, because you have trained them.

You've rewarded them by answering them.

The first is the pace of change in AI.

And I think it's really important to just level set on the type of acid reflux that even people in AI are feeling.

I started in AI almost 20 years ago, and the pace of change is even faster than I would expect and that people in the field are expecting.

Education heavily concerns me.

The fact that companies have not yet leaned in and skilled up their employees, that's a really big one.

The fact that parents have not leaned in to have these open conversations with their children about the risks of

about mental health risks, about over-reliance, about misinformation, about cheating on schoolwork.