AI as an Autonomous Decision Maker

I think the most important thing to understand about AI is that AI is

not a tool. It is an agent. It's the first technology in

history that can make decisions and invent new ideas

by itself. Even something as powerful as the atom

bomb could not decide anything by

itself. All the decisions were made by humans.

Now we've created something which potentially can take power away

from us. At present, it starts with very small things.

Like for instance, uh, there was, uh, an experiment when

OpenAI developed GPT-4, like two years ago.

They want to test what can this thing do, so they gave

it a task to solve CAPTCHA puzzles.

The CAPTCHA puzzle's like when you go online

and you want to access your bank or

whatever, and they have this riddle that you have to solve, an image

that you have

to say what are the twisted words and then letters to make sure you

are not a

robot.

Yeah, it's tough.

You're a human being.

I've taken the test. It's tough.

Yeah.

Is that a street light?

No.

Is that a bicycle wheel? I don't know. Let me do it again. Refresh.

And it's, it's really difficult for GPT-4.

GPT-4 could not solve the CAPTCHA.

But what GPT-4 did, it accessed Taskrabbit,

which is an online site where you can hire humans to do

different things for you, and it asked- ...

a human to solve the CAPTCHA for it.

Mm-hmm.

Now, the human got suspicious. The human asked, "Why do you need

somebody to solve CAPTCHA for you?

Are you a robot?" It asked directly, "Are you a robot?"

And GPT-4 answered, "No, I'm not a robot." "I

have a vision impairment, which is why I can't solve the

CAPTCHA- ... so I need your help."

Oh.

So the fully... So the, the truly evolved human

is not somebody who's smarter, it's

just somebody who gets somebody else to do the work for them.

Smart.

Yeah.

Scary.

Very scary.