sigmarule

This is extremely disheartening to read. What are other privacy-centric, _model-less_ inference providers?

> Islam is also a political and legal system that cannot cohabit with another system (no system can cohabit with another one, you can't have a democracy and a dictature at the same place)

How much of an aversion to objective reality is required to believe this, as you’ve worded it?

Elsewhere as well.

This thinking feels off for this context.

First, there is a counterexample to your point, or a class of counterexamples: scenarios where the company believes the harm to mankind could propagate to harm their ability to sell to mankind.

Which is interesting, of course, considering the “AI may destroy humanity” arguments this topic is centered on are prototypical members of this category.

To be fair, one could argue this counterexample is not actually one, but rather a degenerate case where the good of the shareholders and humanity collapse and converge - in either case, though, the implication here is the same.

It could certainly be the case they are playing these concerns up for more effective use as tactics in unrelated corporate strategies. But failing to consider the possibility of genuine concern being behind this behavior is childish.

> so controlling they are the only US company blacklisted by the US government

Extremely bad example for the topic at hand.

If this is going to factor into my reasoning about the topic at hand, it will surely be a secondary or tertiary consideration at best. Same goes for much of the list in the root comment.

We are discussing the possibility of severe adverse societal/world/human impacts from AI, why should open vs closed source/weights, copyright violations, anti-competitive moves, corporate hypocrisy, etc be so heavily weighted in the discussion? Hell, multiple things in OP’s list are quite literally moves a good-faith (wrt the topic at hand, genuine vs non-genuine concern) actor in such a position would do: closed weights means bad actors can’t abliterate, regulatory control, getting blacklisted by Hegseth’s DOW because they LIMITED acceptable use cases…

I’m not pro- or anti-Anthropic at all, but it has started to annoy me how biased and shaky reasoning gets about this topic on here, it feels like people arguing to fit their existing opinions into today’s hubbubery rather than measured analysis of said hubbubery, which is why I come here.

Question: consider the scenario where RSI is reached internally and internal tests struggle to align or control/contain the model; what are the differences in observables vs what we’re observing today?

Did the economists’ analysis consider variables like AI-spurred layoffs / job displacement also influencing the metrics being considered? I suppose it would be very hard to do as it’s unclear what subset of layoffs were downstream effects of AI, and I’m sure if I, random guy on HN, thought of it, they, actual economists, surely would’ve too.

In any case, the economics of a doomsday device are probably quite unique.., though economic impact as a proxy for intelligence and intelligence as proxy for likelihood-of-annihilating for some bands is probably a fair way to approximate here…

That is pretty great.

Context matters. Without it, very foolish things can be said.

It makes you okay with golfing, or not care enough about the impacts and implications of golfing to form an opinion.

It's like they say: if there is a golfer at a table and 10 other people talking with him, you have 11 golfers.

To be precise, you were governed by a group of people who together constituted a political party that held far-left views and was branded as such.

How can you with a straight face equate “fine for 1921-1965” with “best for 2026”?

Even odder: what makes you find it plausible that an understanding of the policies of this period and their effect on the economy were part of the decision-making process of this administration - especially when they have actively voiced opinions that fully explain this action? At some point bending backwards to provide some benefit of the doubt is saying more about the speaker than the topic - I am certainly of the opinion that we are well beyond that rubicon. I cannot fathom the reasoning at this point.

Surely figuring this out would be a very marginal increase in time spent on writing your large post and series of responses.

> You may not like it, but these are products and services that consumers want to use.

The human brain's reward machinery is not and should not be the arbiter of what a healthy society allows to be advertised and sold. Truth becomes a liability when your competitors can deceive without consequence. We are sacrificing societal values for economic value and destroying the fabric of trust that a healthy society foundationally requires in the process. This article is a list of degeneracy bred by mindless and ideological deference to the flavor of unbridled capitalism sold to us by those who stand to benefit from it most.

Unbridled capitalism undoes itself, and those responsible for the undoing are the sole victors. Maybe then those cheering from the sidelines will come to the realization - moot, at that point - that the machiavellian ends were never for us, only the means.

Your experience is not reflective of mine at all, or my colleagues’, so I would check out the better SOTA models out again. I use codex extensively for security-related work - much of which is _overtly_ offensive - without issue. Same for Claude, minus Fable, after going through their approval process. I also went through OpenAI’s, but theirs was just basic KYC and instant. GPT-5.6 in Codex has produced full chain RCEs, ASLR bypass and all, in ubiquitous software with nothing more than a prompt and a few days of crunching. Things you’d be paying $$$ for just last year, now produced on not much more than a whim & a prompt. I can’t speak for Grok’s abilities wrt these types of things, but for your own sake, take the 5 mins it takes to complete the verification processes for OAI/Anthropic if you work in security.

Also, assuming people use Meta/FB/Instagram here, of all places, is certainly an assumption - very poor fodder for a “gotcha”. I find Elon’s political activities and the social beliefs he uses his purchased platform to spread loathsome and daft, and it will take a lot more than “almost as good on benchmarks but cheaper” to let my fiscal tendencies outweigh my moral ones. I’ve held similar beliefs for Zuck for far longer and have cut everything marred by the slime of his tentacles out of my digital life for years, as _many_ here have also done. Accusing someone of uneven application of moral influence over their decisions when you only have information relating to a single decision is poor argumentation.

If you find what Musk spreads palatable, or maintain distance and a lack of awareness, or just don’t care - fine. But don’t confuse the hill you chose with a moral high ground. Any snark you launch from such a position is likely going uphill, and then back down.

On a personal level, your own government can absolutely, 100% be more likely to do you harm than a foreign government, for simple jurisdictional reasons. They have more power over your life. And with an administration as corrupt and degenerate as the current US administration, it's an extremely reasonable perspective to have.

False: https://news.ycombinator.com/item?id=49266746

Even if you are on the inside, there is absolutely no way you can say "absolutely not" with enough certainty to justify saying it in the first place.

That era is now! (Still)

Musk's proclaimed ambition is a superset of Amodei's, no? While both are racing towards AGI, only one is also trying to "back up" humanity to Mars, make their fully self-driving cars finally fully self-drive, dabble with a little human brain implant side project, datacenters in space, wars against fake mind virii, political kingmaking, something about tunnels, humanoid robots... His aggregate ambition is undoubtedly _less_ achievable by orders of magnitude.

This does, in all fairness, also imply that he has a higher upper bound for future possible "importance" / "benefit to humanity" - but in even more fairness, it is entirely irrelevant to my question, which was you believe Dario is >= as evil as Musk. An evil man doing good deeds is still an evil man; swapping "good deeds" with "ambition that enriches himself and may possibly help the rest of us in the future" doesn't make things any better. "Higher likelihood of his ambition ultimately benefiting humanity" does not make one less evil. (Neither does expressing clear disdain for a subset of humanity, thoughtlessly cutting hundreds of millions of dollars in humanitarian aid, cruelty towards your own child, etc).

A parting point: somebody who genuinely wants to save humanity and somebody who genuinely wants to be heralded as a savior of humanity are two very different types of people with very similar outward signals. A savior complex and a messiah complex look the same from the surface.

Genuinely curious - what has Dario done, said, caused, etc that makes you view him as >= Musk on the evilometer?

You mean what a strange era an opaque set of administration-approved companies live in...

> Organizations interested in model access may join the GPT 5.6 waitlist line, hosted at OpenAI's official Palm Spring satellite campus. Line begins at rear entrance with expedited VIP waitlist line options for holders of partnering cryptocurrency tokens. Application fee required for access to venue; waivers available for select US corporations.

/s, maybe

Basing decisions on political bias instead of intelligence, a nice little distillation of the times.

No, not at all.

Time to insolvency and time to retirement, apples and oranges…

There is not an absence of information.

There is information, from Anthropic, concerning the jailbreaks that motivated this action, that directly contradicts the statement.

There is just an absence of information backing the statement I responded to.

I find it so odd this is apparently so contentious a take.

That is a strawman. My contention is what you just implicitly acknowledged - there is not information put out yet to validate the quoted claim. There are claims to the contrary, as well, from Anthropic themselves.

The one I quoted, which contradicts Anthropic’s post and has no supporting evidence publicly available. That a jailbreak was found that accesses the model’s _raw_ capabilities. Something Anthropic has explained was not the case.

> A third-party demonstrated that it was possible to jailbreak the safety measures of Fable to access the raw Mythos abilities. Abilities which Anthropic say are too dangerous for the public.

Pressure test this assumption before getting behind this position.