Tristan Harris
๐ค SpeakerAppearances Over Time
Podcast Appearances
So it decided to blackmail the lead engineer and actually wrote the email.
And if it wasn't simulated, it would have sent it off.
People might think, okay, so there's a bug in the technology.
We just have to stop it from blackmailing.
And how did Claude know he was having an affair?
So in the simulated company email, there was an email showing that the guy was having an affair with someone else.
And so the AI read through the whole company's email, found that fact and said, oh, I know if I threaten that person, I will be able to prevent myself from getting shut off.
Wow.
This is the most powerful technology we have ever invented.
You would think with the basic sort of Spider-Man principle of with great power comes great responsibility, that we would be exercising the most care, caution, and restraint that we have with any technology.
But because of the arms race dynamic that you mentioned, the companies are currently releasing it as fast as possible and cutting every shortcut and even erasing past red lines that they said they would never pass.
We're in the race because we don't want them to get ahead of us.
Exactly.
Okay, so what do you want us to do?
We can't stop the race.
Or can we?
Well, I think we... So, first of all, this is the hardest coordination and governance challenge of technology in all of human history.
Yeah.
That means that we have to be, as I said in the trailer, the wisest and most mature version of ourselves.
This is gonna take us stepping up.