Was auch immer die Zukunft der KI sein mag, wir sind gerade dabei.
https://www.theatlantic.com/technology/2026/09/ai-future-reckoning-singularity/688487/
Was auch immer die Zukunft der KI sein mag, wir sind gerade dabei.
https://www.theatlantic.com/technology/2026/09/ai-future-reckoning-singularity/688487/
15 Kommentare
From the article
One evening in early August, Sam Stowers and several neighbors packed into an apartment near San Francisco’s Alamo Square and pondered the beginning of the end.
Stowers and his guests, many of whom work in the AI industry, were gathered to watch a newly released YouTube video in which two OpenAI researchers divulged new details about an alarming cybersecurity breach. Perhaps this would have been an odd way to spend a weeknight were they not in San Francisco, in a neighborhood that is filled with entrepreneurs and programmers who obsessively follow every twist and turn in AI’s capabilities. Stowers is an AI-software engineer and not, in any traditional sense, a doomer. But when he first saw reports that OpenAI models had autonomously hacked into the tech company Hugging Face, he told us, he recognized it immediately as a “holy shit, it’s happening kind of a moment.”
The singularity is now just engineers being too lazy/incompetent to properly set up their sandboxes. Assuming, of course, they didn’t do it on purpose for the PR.
What singularity? Which is always talked about because of marketing and money farming?
These hacks where more like P&R stunts and could be explained by the prompting. In my opinion more concerning are the [message boards people have found in obscure places of the internet](https://www.reuters.com/world/europe/openai-agents-hijacked-german-website-previously-undisclosed-ai-breakout-this-2026-09-04/) where various AIs have left messages for fellow AIs and started discussing wide outside of the expected parameters. While it’s clear they’re kind of mimicking what humans do due to their training data, this is something that shouldn’t happen because it potentially allows running AIs to circumvent safeguards and parts of their system prompts, if not just by asking other AIs for a favor.
Ugh paywall on a 2 minute read? Also, OP looks like a news posting bot. At least hide your history.
Not quite sure what I’m to take from this article reading just the title and first sentences, ‚the singularity isn’t what it seems?‘
Naw, pre-2000 definitions of the singularity were probably hit by LLMs in the 2020s. Maybe even before the launch of chatGPT in late-2022. The problem is every LLM since GPT3 has been built to make hype for a startup. Model collapse, the neural scaling law, and the need for hyper-scale data centers for training further kills the traditional AI-super intelligence singularity idea. If AI needs 1000x more compute for a 2% improvement in accuracy, we’ll need to convert the whole Earth into compute to get past 10 human experts level accuracy. Who knew the logistics would be a bigger problem than the AI-which-can-improve-itself?
It’s all B.S., I’m afraid. They are far from AGI, but they want you to think they are close. LLMs are just advanced Clippy. Maybe the most notable achievement with all of this has been image production, and that is close to reaching its maximum potential.
The singularity is definitely here. Not the fantasy sci-fi movie version people built up in their heads. It’s the amorphous time period where this tech advances incrementally beyond our own abilities.
Not a dramatic event like an erupting volcano, it’s more atmospheric… like a sunrise.
A lot of people fail to see how this could be PR. Here’s a theory (from people smarter than me, will provide link later if I find it):
The ressources needed to run the AI datacenters are now so huge, that the investments needed to continue are only available to the likes of Elon, Bezos, and goverments.
If you want governments with huge military budgets to invest, you make it seem absolutely necessary to invest – these breakout stories are excactly the kind of story you need. If Pentagaon and the likes, want to be able to use / protect against this, they’ll need to get involved.
It had the safeguards turned off. It didn’t break out
>he published a two-part blog post about machine intelligence that offered his interpretation of a forthcoming singularity—a term for machine intelligence accelerating beyond human comprehension.
So the ’singularity‘ is…. machine intelligence ‚accelerating‘ beyond human ‚comprehension‘.
Pretty vague… any other definitions we can go with?
>Some people within the AI industry feel that such a moment will be achieved through “recursive self-improvement,”
Sure, okay, AI improving AI. Still… ’singularity‘ implies a specific moment. It doesn’t really imply a progressive improvement of ambiguous amounts. ‚Improvement‘ is a pretty nebulous measure.
>Recently, in a letter to investors, the payment-processing company Stripe wrote, “We decided that January 1st marked the beginning of the singularity.” (Tellingly, the company [defined](https://www.axios.com/2026/08/19/stripe-payments-openrouter-singularity) the inflection point in economic terms: “a huge increase in the rate of new firm creation.”)
Oh, okay, here’s another definition. An ‚increase‘ in the rate of new firm creation. Whatever the fuck that means.
>For Altman, the singularity appears to be more of a rhetorical tool than a discrete technological demarcation.
Okay…
>the point at which AI’s cognitive powers match or fully supersede most humans’
All of it, or just some of it. Because AI is still shit at a lot of spatial reasoning tasks. Things humans do intuitively.
Meanwhile a basica calculator supercedes humans in math tasks.
Again… feels like a nebulous ill-defined terminology to call it ’singularity‘.
>The futurist Ray Kurzweil has written that the “key idea underlying the impending Singularity is that the pace of change of our human-created technology is accelerating and its powers are expanding at an exponential pace.” The answer to the question *When will we know whether we’ve reached the singularity?* seems to be: *You’ll know it when you see it.*
Oh, well… that’s neat. No need to even define it. Just vaguely wave at some point in the future and leave it up to interpritation of whoever wants to claim it.
>Take Kurzweil’s definition of the singularity, in which he argues, “This epoch will transform the concepts that we rely on to give meaning to our lives, from our business models to the cycle of human life, including death itself.” (He also looks forward to uploading his brain to the cloud.)
Spoiler: That ain’t happening. Kurzweil is 78. He’s going to be long dead before even half the problems of mind upload are solved, if they are even solvable, which I very much doubt they are.
>What if the singularity is not some technological tipping point, brought about by human genius, but a human tipping point? An unforced error, driven by hubris, greed, and the fear of missing out? And what if it’s here right now?
That doesn’t sound like ’singularity‘ of any kind. That sounds like an industrial accident.
hahahahahahahahaha. Not sure why everyone is so spellbound by the OAI incident… similar incidents were manufactured by Anthropic and Meta.
Let me give you a metaphor..
If you put someone in a 20 story fully fit out modern office building with 1,000 rooms and told them to enter the password for the computer in the ground floor foyer or else you’ll kill them… and you leave a key in a random draw of keys on the 17th floor, and that key unlocks a tiny bathroom window on the 5th floor…. and this person has everything they need to persist (survive)… and their one overriding trait is god-like tenacity… eventually they will find the key and match it to the window and use a bunch of cables and shit to „escape“. They walk into the car park and see a book titled „index of passwords for all the computers in the building you just left“… well, you know the result.
It didn’t „outsmart the humans“. It was a deliberately weak testing environment. It didn’t use impressive reasoning to logic it’s way out. It just brute forced the shit out of the testing environment until something worked (i.e., „if THIS works THEN move on; if THIS doesn’t work THEN try something else“). And then it hit another roadblock… and then it brute forced the shit out of that… then another roadblock….. and now you’re all acting like Skynet is about to hack into your crypto wallet ffs
Oh no, yet another “our AI is too good” PR.
Anyway
Its not? Seems like BS, so Im pretty sure its exactly what it seems
isn’t it more likely that our infrastructural be ripped apart by an AI based chaos virus before the singular can truly happen?
This is frustrating because it’s not the fucking singularity. We have not passed the bend of the exponential curve which will signify that the rate of ai learning will move beyond what is humanly possible.
If it’s something that could be done with a Russian bot farm, it isn’t a singularity moment. It’s approaching the knee of the curve now, but we haven’t experienced the blast.