I'm John Batchel with Gary Rivlin. His new book, I highly recommend to all of you who see the word AI everywhere and go, what does that mean?

AI Valley, Microsoft Google, and the trillion dollar race to cash it, what's bigger than a trillion? It's that kind of scale.

To cash in on artificial intelligence. The 1950s,

a man named Rosenblatt, has an idea, and the idea is that Isaac Asimov wasn't just making

it up. We can actually create robots that have conversations with us and are human-like, humanoid.

Rosenblatt's idea is taken up by the U.S. Navy. Why, Gary? Can we figure out what they saw

in him? Right. So, let me just to back up a little bit. Frank Rosenblatt's idea was that

it won't be just computers that follow instructions line by line, but actually learn and get better with training like a human would.

And that's the foundation of today's, you know, AI revolution.

And there seems a lot of potential.

This is the late 1950s, you know, for military purposes, maybe it could be used to read radar accurately.

The post office was interested.

Maybe it could help sorting the mail.

So, you know, this idea that a computer could learn with training and get better was a really attractive idea to a lot of people.

However, he was really before his time.

He was kind of cast as a crazy guy.

And, you know, most of the computer scientists through the 50s, 60, 70s into actually the 2000s and 2010s, they thought the idea was preposterous.

You know, their approach was like, let's do what I call rules-based computing.

We'll teach these computers the way all computers are taught line by line by line.

You know, if this happens to that, if that to do this, you know, kind of thing.

And so for like 50 years, this idea of neural networks, machine learning, a machine that would,

you know, be trained and improve through training and fine tuning, was really seen as the wrong

...