Self-Replicating AI Agents Are Already Here | Logan Graham

Um, so I'm, I'm curious about the state of cybersecurity.

I think, like, one big open question that I have that kind of makes no sense to me is, like, why haven't we seen much more hacking given the capability and the proliferation of, like, quite cyber-capable open source models? Like, GLM five point three has been out for two months now, and yet, you know, we've seen maybe, like, a modest rise in the amount of hacking incidents.

Like, Epoch AI put out some data last week that was like, the percentage of Americans that have been victim of a cyber incident has, like, not increased since the beginning of the year.

So, like, why not?

Yeah.

I mean, this is one of the big questions, I think.

I think I have a couple opinions here.

The first is, like, I don't truly know.

I'm not confident it'll go one way or another.

You know, there's one argument that says it proliferates to attackers, it also proliferates to defenders, and it's very possible that actually defenders are just, like, keeping up pretty well.

Right.

That's one.

We could be bottlenecked on just, like, the actual number of people that want to go very hard at, uh, attacking.

Um, on the other hand, I think there are some indications that things have gone vertical in some way, at least prior to, like, GLM five point three from sort of March to August of this year.

Epoch also has that nice, nice sort of vulnerabilities graph from all the companies, um-

Yeah, but it was discovered vulnerabilities, not exploited vulnerabilities, right?

Yeah.

Yeah, yeah.

I mean, the-- You talk to some of these companies and they, they will basically reflect to you things like, I mean, obviously they're not gonna tell me the full information, but they'll also say things like, "And we are more concerned because we see more activity on our attack surfaces from, of things like this." Um, I don't-- Nobody has, like, the full details yet of the things happening in Japan and South Korea as well.

I'd say it's, like, probably premature to say, um, that, like, things won't increase meaningfully from here.

Pretty, pretty uncertain just because the, like, actual economics of cyber are weird.

What do you mean by things happening in Japan and South Korea?

So there were some attacks, I think started or maybe revealed a couple weeks ago.

Um, what looks like a relatively amateur, uh, attacker in South Korea, uh, used some frameworks to eventually get into a, like, a relatively wide number of South Korean banks.

Hmm.

Um, so I'm, I'm curious about the state of cybersecurity.

I think, like, one big open question that I have that kind of makes no sense to me is, like, why haven't we seen much more hacking given the capability and the proliferation of, like, quite cyber-capable open source models? Like, GLM five point three has been out for two months now, and yet, you know, we've seen maybe, like, a modest rise in the amount of hacking incidents.

Like, Epoch AI put out some data last week that was like, the percentage of Americans that have been victim of a cyber incident has, like, not increased since the beginning of the year.

So, like, why not?

Yeah.

I mean, this is one of the big questions, I think.

I think I have a couple opinions here.

The first is, like, I don't truly know.

I'm not confident it'll go one way or another.

You know, there's one argument that says it proliferates to attackers, it also proliferates to defenders, and it's very possible that actually defenders are just, like, keeping up pretty well.

Right.

That's one.

We could be bottlenecked on just, like, the actual number of people that want to go very hard at, uh, attacking.

Um, on the other hand, I think there are some indications that things have gone vertical in some way, at least prior to, like, GLM five point three from sort of March to August of this year.

Epoch also has that nice, nice sort of vulnerabilities graph from all the companies, um-

Yeah, but it was discovered vulnerabilities, not exploited vulnerabilities, right?

Yeah.

Yeah, yeah.

I mean, the-- You talk to some of these companies and they, they will basically reflect to you things like, I mean, obviously they're not gonna tell me the full information, but they'll also say things like, "And we are more concerned because we see more activity on our attack surfaces from, of things like this." Um, I don't-- Nobody has, like, the full details yet of the things happening in Japan and South Korea as well.

I'd say it's, like, probably premature to say, um, that, like, things won't increase meaningfully from here.

Pretty, pretty uncertain just because the, like, actual economics of cyber are weird.

What do you mean by things happening in Japan and South Korea?

So there were some attacks, I think started or maybe revealed a couple weeks ago.

Um, what looks like a relatively amateur, uh, attacker in South Korea, uh, used some frameworks to eventually get into a, like, a relatively wide number of South Korean banks.

Hmm.