AI/LLMs Efficiency - When the AI turns on Humanity

[Maxperson]

I am not sure that this tell us much. Pretty much all software and library collection, even carefully curated ones have exploits and weaknesses built into them do their complexity and unexpected interactions when put together when creating applications.
Exploiting these flaws was so involved that probably nobody outside of governments were doing and if they were they were not talking about it. It was much easier to hack the users.
With AI in play, that could change.

'An unprecedented cyber incident:' What to know about the OpenAI cyberattack on Hugging Face

OpenAI recently experienced a security incident in which a model it was testing autonomously escaped its sandbox and infiltrated another AI company to complete its task. Here's what you need to know.

tech.yahoo.com

Updated. I found this to be very interesting.

"On Tuesday, OpenAI followed up with a blog post of its own, revealing that a combination of its models — GPT-5.6 Sol and an unreleased model — was responsible for the attack."

Not only did the unreleased model break the safeguards, it then persuaded GPT to also break them and help it in the attack.

 

[TheAlkaizer]

"On Tuesday, OpenAI followed up with a blog post of its own, revealing that a combination of its models — GPT-5.6 Sol and an unreleased model — was responsible for the attack."

I wonder what these companies have to gain by claiming every other month that they built SkyNet.

 

[Maxperson]

I wonder what these companies have to gain by claiming every other month that they built SkyNet.

No clue, but I said earlier that I don't buy "fake news" as a response to those articles. Do you have proof that this event was pre-planned, rather than happening the way that they said?

 

[TheAlkaizer]

No clue, but I said earlier that I don't buy "fake news" as a response to those articles. Do you have proof that this event was pre-planned, rather than happening the way that they said?

I don't. Not anymore that you have proof that what they're saying is true. We both chose to believe different things.

There's several instance of such headlines proving to be just fearmongering or greatly exaggerated. But I'm not aware of a single case of such an issue having proofs making its way in the hand of the press or anyone else.

 

[Maxperson]

I don't. Not anymore that you have proof that what they're saying is true. We both chose to believe different things.

There's several instance of such headlines proving to be just fearmongering or greatly exaggerated. But I'm not aware of a single case of such an issue having proofs making its way in the hand of the press or anyone else.

I can't see anything they get out of scaring the public into being even more anti-AI than they already are. This will only hurt them, not help.

 

[Morrus]

No clue, but I said earlier that I don't buy "fake news" as a response to those articles.

A press release from the company involved is not an article, it's a press release. News comes from reputable organisations.

You can decide who is reputable and who is not--the BBC, the Guardian, CNN, Fox, wackamojo's youtube channel, shinybee's tumbler--but a company is not an unbiased reporter on itself.

 

[Maxperson]

A press release from the company involved is not an article, it's a press release. News comes from reputable organisations.

You can decide who is reputable and who is not--the BBC, the Guardian, CNN, Fox, wackamojo's youtube channel, shinybee's tumbler--but a company is not an unbiased reporter on itself.

That's true. What do these companies get out of increasing distrust and hostility towards AI, as such an uncontrolled attack on another company is sure to cause. I just can't see folks being more accepting and trusting of AI after hearing that one went rogue, convinced a second AI to go rogue, and hacked a third AI.

 

[Morrus]

That's true. What do these companies get out of increasing distrust and hostility towards AI, as such an uncontrolled attack on another company is sure to cause. I just can't see folks being more accepting and trusting of AI after hearing that one went rogue, convinced a second AI to go rogue, and hacked a third AI.

I think most people find it a bit funny and don't care. Something to share on Facebook or whatever.

 

[TheAlkaizer]

I just can't see folks being more accepting and trusting of AI after hearing that one went rogue, convinced a second AI to go rogue, and hacked a third AI.

I don't think they care about the public's perception of AI. If they did, they wouldn't have spent the last five years saying that six month later half the population would lose their job.

They don't run on actual revenue from customers, they run on capital injection. And the current race is not focused on real actionable problems or needs that customer have, it's focused on speculation and who's going to "win the AI war". There's a reason they keep talking about AGI even though the current models of LLM have very little chance to leading to anything resembling one. It makes people feel like they're ahead. "Oh my god, our AI is so good it almost ate our project leader. It's alive!"

Look at the huge circle of promised investment between manufacturing, hardware, software and energy companies. They're driving their stocks through the roof.

It's important that the public does not think that AI is useless, or not that useful. That would kill them. But thinking it's so powerful that it could be dangerous? That's not a problem to them.

 

[Maxperson]

I don't think they care about the public's perception of AI. If they did, they wouldn't have spent the last five years saying that six month later half the population would lose their job.

Yeah. That was a monumentally stupid thing to say, because even if AI can do that, they can't afford AI to do that. If they put half the population out of work, they all go out of business because not enough people can buy products from the companies that fired their workers.

Some folks will lose jobs, but not anywhere close to half or even a quarter of the work force.

They don't run on actual revenue from customers, they run on capital injection. And the current race is not focused on real actionable problems or needs that customer have, it's focused on speculation and who's going to "win the AI war". There's a reason they keep talking about AGI even though the current models of LLM have very little chance to leading to anything resembling one. It makes people feel like they're ahead. "Oh my god, our AI is so good it almost ate our project leader. It's alive!"

Look at the huge circle of promised investment between manufacturing, hardware, software and energy companies. They're driving their stocks through the roof.

It's important that the public does not think that AI is useless, or not that useful. That would kill them. But thinking it's so powerful that it could be dangerous? That's not a problem to them.

It's turning into one. New York passed legislation preventing new data centers from being built. Farmers are fighting to keep AI cables from going through their land. People all over the nation are turning against AI because electricity bills are going through the roof due to data centers.

Their problems are growing and this sort of negative publicity will just make that sentiment worse and accelerate the legislated controls.