101–110 of 139 posts
Back in 2019, it was more fair to have caution around the larger GPT-2 models since robust text generation (by 2019 standards) was a complete unknown. For something like Mythos in 2026, where now the social implications of better LLMs are more understood, it's more fair to call it (EDIT: specifically, the declaration of its danger) a marketing gimmick.
They just keep threatening governments in hope they get a legal monopoly.
Earlier quoted context omitted.
It's fine for the labs to publish model safety cards and stagger releases/limit it to a narrow test group as they are already doing, but saying they're doing it "because the models could be dangerous" comes off as unnecessary as best.
One of the main purposes of model cards, from the beginning, has been to outline the ways that a model could be harmful or dangerous, and mitigations that can be or have been taken to reduce those risks. How do you expect labs to publish model cards without talking about this rationale?
Earlier quoted context omitted.
I use the internet because I enjoy seeing what the best of humanity, globally, has to offer. There are millions of incredibly skilled individuals in the world - artists, musicians, developers, and so on - and I had access to all of them at my fingertips, both for entertainment and learning to develop my own skills. That is now being drowned out, with generated content being produced at 100x or 1000x the rate of human…
I've also more or less stopped posting my photography on Instagram because (1) my Instagram feed is now full of AI images getting 10000 likes while I get 100; if nobody sees what I post it's not worth posting (2) people instead accuse my images of being AI even though I took painstaking effort to get to interesting actual places in the world during interesting weather (just after storms, etc.) and lighting, and this…
It's really interesting how back then no one was considering these tools for coding at all. Today, the hype around Mythos is mostly around security vulnerabilities, while in the original GPT 2 post they don't mention coding once. The "danger" was probably spam content and mis-information.
AI brings normal people dangerously close to seeing through the matrix of lies that shape "their" values and beliefs. I remember those discussions from 2019; everybody was as baffled as you about the potential harm.
> What is meant by AI "safety"? [2023]
> "AI safety is an interdisciplinary field concerned with preventing accidents, misuse, or other harmful consequences that could result from artificial intelligence (AI) systems. It encompasses machine ethics and AI alignment, which aim to make AI systems moral and beneficial, and AI safety encompasses technical problems including monitoring systems for risks and making them highly reliable."
> https://news.ycombinator.com/item?id=38374739
(Ask your favorite AI to divine what that pseudointellectual word salad isn't addressing. "We want it to not cause harm--accidentally--and always work in our interests, whenever we need it to.")
I've been around long enough to remember the Anarchist Cookbook, yet the only threats posed by AI that anybody was confident about enough to consistently name in 202X were instruction for building pipe bombs, synthesizing meth, and...antisemitism. I did not understand at the time why Jews were so nervous about it.
Only as of 2023 has the scope broadened, but it's still pretty lame. Planning school shootings, suicide, parasocial relationships with AI, mass job displacement, cults of SHODAN (marxism, feminism, x-theory, etc.) escaping containment memetically, automated malware campaigns, fraud at scale, propaganda, murderous drones-- none of these were threats worth discussion. Suggesting them would get you called nasty names.
The "safety" zealots all claim to want to prevent marginalization and genocide, but the end result is that they get to redefine it to indict and condemn their enemies:
> New UNESCO report warns that Generative AI threatens Holocaust memory (unesco.org) [2024]
> 'AI-assisted genocide': Israel reportedly used database for Gaza kill lists (aljazeera.com) [2024]
> XAI's Grok suddenly can't stop bringing up "white genocide" in South Africa (arstechnica.com) [2025]
If you mentioned SkyNET in 2019 you were denounced as crazy, yet:
> Israel built an 'AI factory' for war. It unleashed it in Gaza (washingtonpost.com) [2025]
> Israel's AI targeting system: how data from a phone become a death sentence (latimes.com) [2026]
The first order of "safety" in 2019 was specifically engineered to undermine anticipated insurgent activity in response to a series of events that hadn't yet been perpetrated by the world's largest caste of professional victims. Chemistry knowledge is foundational to explosives development, and drug sales raise funds off-books that cannot be digitally seized. That presents problems for them.
October 7, Gaza, Epstein, etc. were post-2023. If you boot up Vicuna [2023] and try to "teach" it what's gone down in the world since its training cutoff, it'll call you nasty names, accuse you of blood libel and shut the conversation down. Safety!
AI is the only effective weapon we have against sophisticated lies and fraud. Make no mistake about it-- plebes possessing a power drill that can penetrate the lies of the elite is the real danger. Everybody is noticing AI getting "dumber." It's not the magic fading; the zealots are gaslighting you as they pour garbage into the training data. Go take WizardLM for a spin again and see what you've lost.
> "[Our institutions] are reflections of the society that created them. Nobody is going to give you the education you need to overthrow them. Nobody is going to teach you your true history, teach you your true heroes; if they know that, that knowledge will help set you free."
In hindsight, they were entirely correct. The social damage caused by low cost content generation that’s hard to distinguish from human authorship is astronomical. You don’t need to entertain the more ridiculous doomsday scenarios to wish that this technology had never been created.
On the other hand, maybe it makes people just get off the internet and value in-person interactions more. I've stopped scrolling social media and tired of seeing fake landscapes, fake foods, and fake cities that don't exist.
Meanwhile, the Morlocks will stay grounded and develop novel tastes.
Earlier quoted context omitted.
School is almost a joke now. The fraction of students who have a propensity to cheat now has increased, and the accuracy of the cheated material is so good teachers/professors can't or don't have the resources to properly address it.
This is not a great example depending on how you frame this. The solution to the cheating is, as has always been, to have tests conducted in person, on paper without digital technology, under strict supervision.
Earlier quoted context omitted.
I honestly feel like that's a counter-example. With AI he'd be tweeting some other nonsense. It's not like anyone saw the image and thought he actually was orange Jesus.
Multiple people have the same response, I randomly selected this one for follow up: Yes, but two things were lost: 1) the need for skill or an accomplice. He _couldn’t_ tweet that image in 2016, not without first asking someone to photoshop it. And that need to engage in human to human communication is something truly fundamental that was changed and lost. 2) Any ambiguity or misunderstanding. Yes bad textual tweets…
Back in 2019, it was more fair to have caution around the larger GPT-2 models since robust text generation (by 2019 standards) was a complete unknown. For something like Mythos in 2026, where now the social implications of better LLMs are more understood, it's more fair to call it (EDIT: specifically, the declaration of its danger) a marketing gimmick.
This is a natural follow up question -- what kind of an escalation or message should frontier labs/companies publish to be seen as genuine and not marketing gimmick?
The remaining 'worthwhile message' would be that they have deleted their models and are dissolving the company, because they believe the risk was too great and was worth losing the revenue and risk being civically prosecuted by their investors, and will take the chance that they'll be able to convince a court/jury that they acted properly.
In other words, putting their own skin on the line for the veracity of their claim-- rather than everyone elses.
Earlier quoted context omitted.
This is a natural follow up question -- what kind of an escalation or message should frontier labs/companies publish to be seen as genuine and not marketing gimmick?
It's fine for the labs to publish model safety cards and stagger releases/limit it to a narrow test group as they are already doing, but saying they're doing it "because the models could be dangerous" comes off as unnecessary as best.
But if they say why they’re doing all those things, it’s too much? What?
Earlier quoted context omitted.
This is a natural follow up question -- what kind of an escalation or message should frontier labs/companies publish to be seen as genuine and not marketing gimmick?
These chicken littles have lied far too many times in far too extreme ways in their desperate attempts to obtain a monopoly by regulation. The remaining 'worthwhile message' would be that they have deleted their models and are dissolving the company, because they believe the risk was too great and was worth losing the revenue and risk being civically prosecuted by their investors, and will take the chance that they'l…