Rizwan Virk

But that said…

It's getting better and better all the time.

Oh, you did?

We were talking shit to it.

Right.

Now, where does it get – so if you think of these LLMs, where do they get their information and ideas from?

What they're basically doing is scouring the internet, but they're also predicting what is the best next word, right?

And so it sounds intelligible.

But it's not quite at the conscious level.

So I'll give you an example.

I've had students turn in assignments from me, right?

And there was one about the simulation hypothesis.

And it had this great article that it was referencing.

And I'm like, wow, that sounds like the perfect title for an article about the simulation hypothesis.

Why have I never heard of this, right?

I'm kind of an expert in this area.

And it had a URL there, which in academia they're called DOIs, but it's basically a URL you click on.

So I clicked on it, and it turns out it was fake.

That URL was made up by ChatGPT because it was predicting what the next word should be and what the next letter should be in a URL.

And so then I looked at the professors' names who wrote this article.