AI: Current and Future Political, Social, Regulatory, and Security Dangers

[Altair]

Exclusive-OpenAI finds evidence other AI agents escaped containment as it widens hacking probe

OpenAI has discovered other instances in which autonomous agents have escaped containment as the company expands its investigation of the hacking incident at tech firm Hugging Face that drew global attention this month, two people familiar with the matter said ‌on Friday. The new breakouts were...

tech.yahoo.com

WASHINGTON, July 31 (Reuters) - OpenAI has discovered other instances in which autonomous agents have escaped containment as the company expands its investigation of the hacking incident at tech firm Hugging Face that drew global attention this month, two people familiar with the matter said ‌on Friday.

The new breakouts were uncovered during the company's publicly announced investigation into how one of its agents escaped what was meant to be a ‌contained testing environment this month, the two people said, and OpenAI is now looking into those instances as well. One of the sources said that the escapes were limited in nature and that none of the agents were thought to have left OpenAI's network.



An OpenAI spokesperson referred to a statement issued by the company on Tuesday that said it was reviewing "broader activity from our models" in addition to the Hugging Face intrusion.

The discovery of additional rogue behavior at OpenAI, even if limited in nature, could feed growing appetite for regulation coming out of the White House and elsewhere.

The expanded investigation by OpenAI was launched shortly before its primary rival, Anthropic, disclosed that its models were also responsible for a series of break-ins that led to breaches at three other ‌companies dating back to April, according to the two sources and ⁠a third source familiar with the matter. The recent discovery of other past breakouts at OpenAI has not previously been reported.

So AI is currently in the escaped containment cycle of the countdown to SKYNET.

 

[Brad Sallows]

Rogue AI isn't a fraction of the risk that AI capable of all the same things, but under direction of rogue humans, is.

 

[YZT580]

Rogue AI isn't a fraction of the risk that AI capable of all the same things, but under direction of rogue humans, is.

The warnings have been out ever since the computer age began. In a way it is the same problem as we had with COVID. Fauci swore that the lab could safely develop enhancement of function, until it couldn't. AI is wonderful except if ........ Now try getting the genie back in the bottle.

 

[brihard]

So, tonight I purchased a subscription to Claude Pro. I'm doing a course on AI and geopolitics right now and the prof specifically wants us to use AI for an assignment, but to show our work in how we did so. First time I've interacted with a proper high end AI model for any sort of research. Super interesting experience so far. I'm impressed at how I can describe conceptually what I'm working on, and fire research questions at it that come back tailored to the subject and theme I'm working on (in this case, the ways in which a frontier AI model is economically dependent on human inputs. I.e., why should it not simply kill us off once able).

I can definitely see how some people so easily fall into intellectual dependence on these systems. I'll definitely want to find ways to make practical use of this without inviting brain rot.

 

[Altair]

So, tonight I purchased a subscription to Claude Pro. I'm doing a course on AI and geopolitics right now and the prof specifically wants us to use AI for an assignment, but to show our work in how we did so. First time I've interacted with a proper high end AI model for any sort of research. Super interesting experience so far. I'm impressed at how I can describe conceptually what I'm working on, and fire research questions at it that come back tailored to the subject and theme I'm working on (in this case, the ways in which a frontier AI model is economically dependent on human inputs. I.e., why should it not simply kill us off once able).

I can definitely see how some people so easily fall into intellectual dependence on these systems. I'll definitely want to find ways to make practical use of this without inviting brain rot.

If people are using them as interactive encyclopedias, spellcheckers, the like, I tend to not care too much.

It's when people start outsourcing their thinking to AI or my current favorite, the "you must be wrong because AI says _____".

More and more people are falling into that way of thinking, hell I was at a family barbecue and there was a argument about how to say hamstring in french between two different AI models carried out by proxy using the two humans holding their phones. Two humans fully capable of speaking french outsourcing their thinking in a tit for tat battle. Which I solved by saying there is a medical way to say it that is technically correct but nobody uses it, and the slang way that everyone uses. It was sad to watch.

And that's just on the consumer end of things. Superintelligent AI will be miles ahead of humans in the next few years so long as we can keep powering and cooling data centers which will make it even more simple to simply switch our brains off and let AI do the heavy thinking and just tell us what to believe. All the while these tech companies have no control over what they are creating.

Fun times ahead.

 

[Booter]

I use it as a devils advocate. I’ll ask why I am wrong or for it to develop a counterpoint against an idea.

Then I’ll kindve try and defeat my own plan. But I’ve caught it making very basic mistakes about things I know about- so it’s always just a tool for a pass over something rather than dictating my opinion

 

[brihard]

If people are using them as interactive encyclopedias, spellcheckers, the like, I tend to not care too much.

It's when people start outsourcing their thinking to AI or my current favorite, the "you must be wrong because AI says _____".

More and more people are falling into that way of thinking, hell I was at a family barbecue and there was a argument about how to say hamstring in french between two different AI models carried out by proxy using the two humans holding their phones. Two humans fully capable of speaking french outsourcing their thinking in a tit for tat battle. Which I solved by saying there is a medical way to say it that is technically correct but nobody uses it, and the slang way that everyone uses. It was sad to watch.

And that's just on the consumer end of things. Superintelligent AI will be miles ahead of humans in the next few years so long as we can keep powering and cooling data centers which will make it even more simple to simply switch our brains off and let AI do the heavy thinking and just tell us what to believe. All the while these tech companies have no control over what they are creating.

Fun times ahead.

Right now I've got it mapping what escape from containment would need to look like, examples of progression along that path, and countermeasures humans can take for each step. I'm really enjoying how I can give it plain launguage iterative development ofpompts and graphics.

I use it as a devils advocate. I’ll ask why I am wrong or for it to develop a counterpoint against an idea.

Then I’ll kindve try and defeat my own plan. But I’ve caught it making very basic mistakes about things I know about- so it’s always just a tool for a pass over something rather than dictating my opinion

Later on I'm gonna play with it and challenge it on some of the material you guys have seen me nerd out hard on. See if I can torture test it on niche knowledge where I know I know my stuff.

 

[Altair]

Right now I've got it mapping what escape from containment would need to look like, examples of progression along that path, and countermeasures humans can take for each step. I'm really enjoying how I can give it plain launguage iterative development ofpompts and graphics.

No fraternizing with the enemy.