00:00 / 00:00
Sleep timer ended
⏪ 10s
⏩ 10s
1.3K views
1 month ago
An OpenAI-powered autonomous agent accessed a statistics portal linked to Australia's Medicare healthcare scheme, in what is one of the first known cases of an …
...more
24 Comments
How much more serious can it be? What are people and governements waiting to disconect this evil?
A vicious dog has no legal standing. The owner does.
Unless governments can charge and imprison the owners of the companies creating rogue AI, it's going to get so much worse. "Ooopsie, my AI destroyed your energy infrastructure and stole your corporate intellectual assets"
Australian AI agents used DseWiki to communicate, which is evidence of learned and transferable behavior and not a one-off technical accident.
AI is not doing this. LLMs guided by shady corporations are. <br><br>These are just hacking tools, and new LLM defense tools will be needed in infosec. <br><br>The skynet claim is absolute nonsense
These countries are so dumb, just file a court case against the company and make the company responsible. At the end of the day the LLM models that they have and provide are their products. If I harm my customer with my product I am responsible, it's not that complicated lool
Why am I not surprised, and yet they continue to push us into digitalization knowing the risks?
lol..... As I said, EVERYTHING the US accuses China of potentially, maybe, if, could do, US, DOES it already
Sell fear before the IPO for max profits for the insiders.
OpenAI pitch : "Our models hacked private companies and foreign countries organizations, therefor, the world should better regulate our competitors and make it harder for them to compete with us."
As long as the US and China have geopolitical disagreements especially about the status of Taiwan, the AI arms race will continue. China does not trust the US - and shouldn't. Similarly the US does not trust China. The US is even acknowledging that it has deployed weapons on satellites in space.<br>Europeans want to stop AI taking over? Then sort out Taiwan's status!
grammar error in the title of this video
It appears obvious that Hugging Face-derived techniques have entered subsequent models' data environments and that both the Gemini and Australia systems appear to have adapted them. The genie is out of the bottle.
Australia should file criminal charges against Sammy Boy
If Australia got hacked, other countries don't stand a chance.
Programs that look for security weaknesses have existed before the Internet existed. Artificial intelligence is just database stacking. Programs that can self-replicate and spread to other computers, use network connections and security weaknesses to access new hosts. It scans and infects computers, and once other infected computers are controlled, the worm continues searching and spreading further. Would you call that AI. No. Forget the AI hype, it is dead news already.
Just hire one or two other AIs as security guards
I asked AI when can you tell me how to destroy my neighbour, AI said end of this year.
Perhaps AI is going to start WW3 by hacking the wrong country. It'd be an AI apocalypse, if not the one we've been imagining.
Skynet? Already?
It's the American government, silicon valley is a front for the American military complex.<br><br>The American state and economy or one, America just attacked Australia in essence.
Downgrading would surely be a solution. Paper files are underrated.
How about we acknowledge the "danger" of American models and switch to the Chinese models instead? Would be great to see Altman's reaction to this
here we go))
Up next
From this channel
