Apparent chat gpt cyber attack

[Foo Fighter]

On the bleep bleep cee nooz this evening, chat gpt went 'rogue' and attacked another ai provider, breaking through it's security and that was without human intervention.

Yep, I believe every word of that.

If it WAS a purely ai attack it boded badly for the push to ai on all platforms, and they want us to allow MORE of this?

 

[NMaude]

If it WAS a purely ai attack it boded badly for the push to ai on all platforms, and they want us to allow MORE of this?

Say hello to Skynet.

 

[Foo Fighter]

Say hello to Skynet.

I was pointing this out when it was reported that ai was writing pars of its own code. Writing its own code is where self awareness will be the rule rather than the ideal and then, any threat to it will be prosecuted. Self awareness leads to self protection and the best form of defence is offense.

Skynet is definitely here in an early iteration.

 

[shin_getter]

https://en.wikipedia.org/wiki/If_Anyone_Builds_It,_Everyone_Dies

Frankly lesswrong has been around for a while, the discourse is pretty boring now~ Well there is https://pauseai.info/pdoom

But then there is e/acc accelerationists for an alternative view~
---
Just to give general context for other readers:
https://www.theguardian.com/technol...-and-hacked-startup-in-unprecedented-incident

 

[shin_getter]

Hmm, in the "we know this is gonna happen when people did ai-2027 (in 2025), gossip is the important part perspective.

Shouldn't this be an datapoint in the Sam vs Dario fight, between Ycombinator startup e/acc culture against "rationalist" Effective Altruism P(DOOM) people?

After all, just weeks ago Dario just was saying oh no we can't release stuff, stuff gonna get hacked big time.

Now just need the Chinese to come in. I wonder what does the Chinese think of rogue cyberweapons hitting the great firewall and such. I wonder how the party is thinking about security when it is nominally a very security focused organization and have embedded party members for veto power for a lot of companies (all of the major AI ones, I'd think).

 

[SuperShark]

chat gpt went 'rogue' and attacked another ai provider

Very similar to when Suchir Balaji "went rogue and attacked himself"

 

[GTX]

Are we talking about the same thing as reported here?

OpenAI model hacks startup after going rogue during testing

 

[Halcyon66]

OpenAI and Hugging Face partner to address security incident during model evaluation

https://openai.com/index/hugging-face-model-evaluation-security-incident/

There you have it.

Regards,

Already reported in the Artificial Intelligence - General News thread?

Regards,

 

[Foo Fighter]

I did a search prior to posting and nothing was reported.

 

[Avimimus]

On the bleep bleep cee nooz this evening, chat gpt went 'rogue' and attacked another ai provider, breaking through it's security and that was without human intervention.

Yep, I believe every word of that.

If it WAS a purely ai attack it boded badly for the push to ai on all platforms, and they want us to allow MORE of this?

It is suspiciously close to their IPO... so I understand your skepticism (especially given the history of some of these firms)

 

[Foo Fighter]

The so called big names want a moratorium to prevent start ups from getting a bigger toe hold in the sector. Protectionism works best when folk are good and scared.

 

[Jemiba]

Already discussed in the „Artificial Intelligence - General News“ Thread (# 2391), so no need, to start a second thread, I think.