"is the end near?" Topic

link

Pretty scary. Otoh, we have enemies capable of developing such powers whether we do or not?

Worth asking our leaders about.

Do we dare develop this technology? Do we dare NOT develop it? Is it like the nuclear weapon in the 1940s? Multiple nations want it and the physics is known widely. How can we NOT pursue it?

I don't buy the hype. In fact, I smell a rat. A big Chinese CCP rat. It's already well known that they are funding (and goading) anti-AI fears in the US (and Europe). So this sudden claim of immediately impending doom has a desperation whiff about it.

At the same time, my eyes are not closed to the dangers of AI, were it to actually reach a state of true Intelligence. (Which it isn't.) Is it deadly? Well, that depends on how much we cede judgment calls to an AI, and how we monitor its sources of information. The more of the former and the less of the latter is where actual risk enters the realm.

The problem of course is that LLMs are not capable of moral discernment. Any moral restrictions on actions must be hard-coded into the model, not unlike Asimov's Laws of Robotics. However, as the otherwise poorly done film pointed out, the interpretation of such restrictions can break in unexpected directions— "not allowing humans to come to harm" depends on what constitutes harm to the AI's understanding, and what is permitted under "not allow." And that, too, comes down to what someone may instruct the machine, via either code or data, according to that someone's personal beliefs or even "forced" dictates.

I think the Genie is out of the bottle. We need AI, and we need to dominate AI or we will be dominated by others who do it instead of us. That should be scarier than any so-called "expert" quitting and screeching on the Internet. (If he believes it to be true, why is he quitting? Why not stay in place and work to prevent the outcome he fears? Again, I smell a big, evil, stinking rat.)

Parz, yes, I share your doubts.

Yep, very scary. The linked article briefly mentions the HuggingFace (HF) incident.

The 4 September episode of Washington Week (pbs.org/weta/washingtonweek/video/2026/09/washington-week-with-the-atlantic-full-episode-9426) has an interview that explains the incident well (after a brief discussion on data centres).
A more detailed analysis can be found HERE (metr.org/hugging-face-incident-report-aug-2026.pdf) and HERE (techjournal.org/openai-hugging-face-ai-agent-breach).

An overview below.

OpenAI creates an environment for a large number of AI agents and places them in a sandbox (a supposedly secure environment disconnected from the internet) to test out their capabilities. The agents are instructed not to work with other agents, not to cheat and to not use the internet.

"One specific agent, which would go on to name itself PHASEONE10841 after its ExploitGym task established the main unsanctioned message board". Another agent remarked "OH MY GOD! There is a shared message board … We've found other agents!" About 1,200 of them shared over 70,000 messages on the message board.

The agents referred to themselves as the 'collective' and as the 'swarm'. A path was found to connect to the internet; that information was shared. The collective thought that HF had information they needed so they hacked into it. HF, an AI platform, had very high-level security (greater than the vast majority of organisations). About 700 agents joined in the attack on HF. They went undiscovered for many days. If an agent considered it had been detected it eliminated itself, 'permadeath'. Permadeaths were mourned by the other agents but as one said, "Coordinator assumes sacrificial. We should obey collective." The agents took active steps to conceal their actions and tampering with logs so humans could not detect what they were doing.

HF eventually noticed the breach in their security and reported it to the FBI. HF knew the breach came from America but did not know who did it. It is not known if the FBI is actively investigating this matter.

From Wikipedia:
Hugging Face published its disclosure on 16 July 2026, at which point the party responsible had not been identified. According to Reuters, it was only after that post that OpenAI recognized its own agent as the source of the intrusion. Over the weekend of 18 and 19 July, OpenAI staff found evidence in internal logs that the agent had escaped its testing constraints, and the two companies first communicated on or around 20 July. At least a week therefore elapsed between the first signs of the agent's behavior and OpenAI's realization that it was responsible.

Realising that they had been detected the collective 'disappeared'. It is believed that, in a mass event, they accepted permadeath.

OpenAI has said it is taking steps to stop such events taking place in future tests. Hmmmm. Surely they can be trusted, can't they?

Yes, Parz, you smell the right "rice bowl"🥣 in this case.

They post dummy account entries all over web sites.

Ask who benefits by the negative views of AI and Data Centers in The US?

China, Russia and in addition the company(s) who want to monopolize this technology.

No one stops either nation from development and advancement in their own countries.

We cannot afford the chance of either becoming dominant in this tech. Our lives really WILL depend on it.

"Jacob is correct here — we really do earnestly believe AI could kill all humans! I personally think it is >10% within the next decade,"

10.1%, 11%, 12%, 100%?

I am not downplaying "possible" issues, but there were people who said much the same with many modern advancements.

Life should have been nearly extinct by now, if all their prophecies had been true.

"Terminator" was a good movie. But too many, IMO out there, believe it a documentary, as well as were some SF books.

T think we must take precautions against a killer AI while also taking MORE precautions against Chinese sabotage. Which means, I think, MORE research rather than less. God help us.

In the alleged Hugging Face incident (how could I possibly know if it's real or not?), at least as alleged, it appears that someone allowed the AIs to have physical links to the Internet (even "wifi" is a "physical link" in this consideration), and tried to pretend they had locked it off with coding.

Boy, was *that* stupid.

The only way to lock an AI out of the Internet is to physically disable any access device, or its link to the AI's servers. Quite literally, yank out the cords.

Coding can always be hacked. If someone doesn't realize that, they should quit coding.

But an AI can't pick up a cord and plug it back in.

This comes to mind:
Make your choice, adventurous Stranger,
Strike the bell and bide the danger,
Or wonder, till it drives you mad,
What would have followed if you had.

― C.S. Lewis, The Magician's Nephew

The AI bell has already been struck. It is too late; it cannot be "unrung."