Will AI Really Kill Humanity By 2030? | Aleksandra Przegalinska

Aleksandra Przegalinska (0:00)

AI does not have to ascend to general artificial intelligence or the super intelligence to be dangerous. Their nascent face and are not as advanced are already causing a lot of harm. If they behave this way, against hugging face or some sort of German wiki site, it's very clear that they could behave in a similar manner towards like a banking system and maybe the Pentagon.

David Lin (0:27)

Has AI reached artificial general intelligence levels yet? Is AI about to wipe out humanity? These are some of the biggest questions we'll be examining with our next guest, Aleksandra Przegalinska. She is the Senior Research Associate at Harvard Law School. She's been on the show before. Check out our last interview and she'll give us the most recent analysis on the most recent updates. Welcome back to the show. Good to see you, Aleksandra.

Aleksandra Przegalinska (0:47)

Good to be back. Thank you so much for the invitation, David.

David Lin (0:50)

So many new developments, so many new AI attacks, cyber attacks, and so many doom narratives in the media from not just economists but the AI researchers themselves. Take a look at this particular one. We're going to start here. This is from the request of our producer who really wants me to show you this clip.

I'm going to forward this is a this is supposedly as human controlled, but it's just to illustrate how advanced the hardware has gotten now. OK, this paired with here is Dario Amode speaking at the UN.

Dario Amodei (1:30)

If managed poorly, I even believe that AI could be a risk to humanity as a whole.

Greg Brockman (1:36)

The Renaissance is a time when new tools, new ideas and new institutions expanded what people believe they could do.

AI can play that role for our time. I believe that people are much more capable than they imagine themselves to be, but have been held back by limited technology that was meant to empower, but instead distracts and often steals attention.

David Lin (1:57)

Okay, Amadei, as you know, wrote a letter last week or two weeks ago warning that AI is moving too fast. We need to put guardrails and safe rails and slow down the pace of development. Why are they coming out? And by they, I mean, Dr. Amadei and many others in the field coming out right now with these warnings. And when the general public who may not be experts in the AI field, hear things like this and see things like this paired with a video clip of Terminator punching or kicking a boxer, people get very scared.

Aleksandra Przegalinska (2:32)

Well, they do. And that is not too hard to understand, I think. Well, with the boxer, I would say this is a very masterful hit in a way, because it touches on so many negative sentiments regarding artificial intelligence. It's very well crafted. If somebody wanted to present a doomsday scenario, they really did so well. But regarding Amadei and all those recursive learning agents that are attacking various systems, we have seen that with hugging face in the summer, but we've seen other attacks as well, one in Germany, a very recent one in Australia. And many people claim that that type of news will keep coming back because there are many attacks that we have not uncovered yet, and we'll get to learn about them very soon. So this is definitely very serious. And well, I have two interpretations, David. One is that it's a sincere sort of sentiment. It's a sincere request to be regulated, to impose some guardrails on a technology that is moving very, very fast and is becoming untamed in the sense that those recursively learning agents are, well, acting in a rogue manner, right? So they are conspiring against humans, they're hacking systems, they're acting against the law.

It is true, though, we have to say, that the prompt is zero for them, both in terms of what Anthropic has been doing and OpenAI was for them to be very persistent and just pursue the reward and avoid all possible obstacles and also, if possible, break the rules, if needed, to get that reward. So they somehow contribute to those behaviors of these agents, but the agents obviously do multiply and act as a mob in a way, and that is concerning. So you could say it's sincere, but there is another interpretation that I see very often, and namely that those biggest technological companies, when it comes to AI, those two pioneers kind of conspire to leave everyone else behind, that it makes sense for them to ask for regulation, because they're on the top of that mountain, and for the other players to kind of keep up with them, there would have to be a situation with no guardrails, because if there are guardrails, they're never going to make it. So when you hear someone like, you know, Justin Huang from Nvidia, what he says is that these regulations are not needed. Quite to the contrary, we have to race, we have to race with China, we have to race with ourselves and with the other tech companies, and he does not buy into that narrative of sudden regulation. So I just wanted to be fair and present these two perspectives. I'm closer to the one that claims that regulations are needed. Currently, if you own a food truck, you actually have to abide by many more regulations than you do when you work in an AI company, which is weird.