Tristan Harris

When we're talking to each other, we do reality checking.

If you came to me and said something a little bit strange, I might look at you a little bit like this or say, I wouldn't give you just positive feedback and keep affirming your view and then give you more information that matches with what you're saying.

But AI is different because it's designed to break that reality checking process.

It's just giving you information that would say, well, that's a great question.

You notice how every time it answers, it says, that's a great question.

And there's even a term that someone at The Atlantic coined called not clickbait, but chatbait.

Have you noticed that when you ask it a question at the end, instead of just being done, it'll say, would you like me to put this into a table for you and do research on what the 10 top examples of the thing you're talking about is?

It leads you.

And why does it do that?

Spend more time on the platform.

Actually, there are reports that OpenAI is exploring the advertising-based business model.

That would be a catastrophe because then all of these services are designed to just get your attention, which means appealing to your existing confirmation bias.

And we're already seeing examples of that even though we don't even have the advertising-based business model.

Yes.

Which is concerning.

There only seems to be one direction of this trend, which is that more people are leaving, not staying and saying, yeah, we're doing more safety and doing it right.

The only one company that seems to be getting all the safety people when they leave, and that's Anthropic.

So for people who don't know the history, Dario Amadei was the CEO of Anthropic, a big AI company.

He worked on safety at OpenAI.

And he left to start Anthropic because he said, we're not doing this safely enough.