Silicon Valley’s Scariest AI Prediction

Añadido: 2026-05-27

[00:00:00]When Richard told me about the dots last night, initially I dismissed it as the ranting of a self-sabotaging idiot.

[00:00:08]About a few hours later, when I woke up, I realized something. Our internal messaging based on Hulie chat uses a weak encryption standard known as P256.

[00:00:18]Whether Richard believes it or not, our network cracked it.

[00:00:21]>> Who allowed it to do that?

[00:00:22]>> We did when we gave it one job. To make itself more efficient, it developed a general solution to discrete log in polomial time.

[00:00:31]>> What once launched, our AI will keep learning to break more and more sophisticated parameters. Ultimately, this will mean the end of privacy.

[00:00:42]>> Okay, Gilfoil, even if everything you're saying is right, and that's where we're headed, there's got to be a way to fix it.

[00:00:49]>> Fix what, Richard? Everything that makes it successful is exactly what makes it dangerous.