Top AI Expert MOCKS Zuckerberg, Jensen Huang For Blocking Safeguards

Welcoming to the show now Roman Yapolski, AI safety researcher. How are you doing today, Roman? Thank you for joining the show. >> Pretty good. [clears throat] >> All right. Uh, incredible. Who wants to start? >> Um, Dr. Gambolski, if you could just give our viewers a little bit of background on yourself and you're

someone who's been sounding the alarm about the existential dangers of super intelligence for a while now. So, just sort of sketch out for people where you are in this debate. Yeah, I'm a computer scientist. I'm a PhD in computer science. My doctoral work was on cyber security, behavioral biometrics. I've been working on safety for bots since,

you know, 2006. Then bots were online bots for playing poker. Back then, I try to detect them, prevent them, but they got smarter since. And I try to keep up, but it doesn't look like I'm going to succeed. Well, I actually [snorts] want to ask about that's a really good place to ask about this distinction between uh

the bots of your and older technology and current AI and where it's going. We refer to these AIs as quote agents. And in a sense that's really accurate because they have been delegated or they've been given quote agency in a way that prior bots and if I'm wrong on this please help me understand it, but prior

bots or prior pieces of technology like the cruise control on your car doesn't have agency. you turn it on and off and if it keeps going despite your instructions, that's an error. But, uh, right now AI, it's not like the online poker bot. It has the agency to sort of make decisions after it's turned on that

could prevent anyone's ability from turning it off because they're acting as agents. So, I know that's kind of a deep question, but could you kind of help us understand the technology, the the way that it's actually built uh to have agency, what that means, and how it's different from past technologies? That's [clears throat] a great question. So