i sort of called this yesterday and the day before. it's not just the military and it's not just contract cancelled, it's throwing anthropic into a fair bit of trouble they might not be able to cope with. why? the modus operandi is pretty consistent: accept my outrageous and unacceptable demands or get crushed. this is not how you negotiate or void a contract, this is how you bully or go for the kill. it actually looks more like the alliances were solidified beforehand and this is a deliberate move to eliminate anthropic. openai stands to gain quite a bit from that, it gets the contract on top, it might eventually get the (not unsubstantial) crumbs of what is left of anthropic's work after the witch hunt and the administration gets its toys and sets the mood for everybody else.

ofc the administration might be miscalculating, that wouldn't be a first, so wait and see.

this is undoubtedly very bad, but we shall see ...

imo the constitutional angle is just trump rhetoric.

however, in an hypothetical case and depending on the relationship (if anthropic ever became indispensable for public infrastructure or function ... think of the postal service, telcom, ...) then it could be considered a state actor (as opposed to a private entity) and as such their tos would (or should) be constrained by the constitution.

again, i'm not saying this is the case, but if trump were to be taken seriously that would be the angle and gp asked. i don't interpret trump as saying "they violate the constitution" but rather "they don't allow the department of war to defend american people as the constitution mandates". which is ofc hogwash.

more like two shopping carts fighting around one looter. they're smart shopping carts!

maybe that was in the tos, i couldn't tell bc i haven't read it.

anyway, maybe so, but what's the point of that hissy fit, then? what if the discussion has already been had, they already have a deal with openai (which ofc would keep loudly vowing their "moral" concerns in public while complying anyway), and the reward is precisely destroying anthropic (which is a close competitor to openai)? the cherry on top, once anthropic is broke and busted, what about making a move on its ip, which is in some ways superior? ofc all this is just wild speculation on my part, but one constant in these affairs is that usually nothing is what it seems ... the other is that it's all scumbags all the way up.

depends. enacting such a system would be quite convenient in the aftermath of an "accident" (e.g., a genocide, but many other scenarios come to mind): "oh, crap, this was totally unintended, we had total confidence in this system, we were assured this wouldn't happen and with bleeding hearts we will ensure this will never happen again!". ok, this is all still stupid, but just a different kind of stupid.

again, this is all speculation, but the moment a government takes issue with these kinds of safeguards i would say that clearly anything is on the table.

tbh he doesn't say they violated the constitution, but literally that anthropic wants the war department to respect their tos, which is .... what anyone would expect? if their tos violates the constitution i don't know but that should have been known a while ago, and anyway that isn't what such social-media-royal-decrees aim to dilucidate, being just public gossip with no legal relevance whatsoever.

yesterday i pondered if they wanted anthropic to comply and shut up or be gone, or just to be gone. it would seem it's the latter.

that's easy: they are "authoritarian regimes", everything they do is fake and evil. we otoh are "virginal beings of light and beacons of freedom", everything we do is real and right. that we are the ones having started most of the wars is just us honoring our sacred duty to the light and to the one true god, you heathen.

we hardly ever had any, not even in ancient times. the lies to justify rome's invasion of gaul were very similar to the lies to justify the invasion of irak two millenia later. it's always the same playbook, it's just that the level and sophistication of noise is exponentially increasing. the basic lies, however, haven't changed and don't really resist basic logic, but still work for the most part. now, our capacity to cause harm does increase exponentially too, so that should tell you where we're headed ...

on the surface, yes. but that might just end up meaning anthropic would not be collaborating ... publicly. after all, they have already collaborated in illegal operations, maybe not willingly but even then with gross naivety: you don't just claim to be a safeguard and then partner with the pentagon and palantir assuming they're white knights.

ofc after these outlandish threats have been made it would be now difficult to maintain a good reputation if these don't materialize. nobody could trust anthropic either. so it looks like they want them either complying and shutting the fig up or gone, or simply just gone. i have to say i welcome the style of these aficionado bullies running the show right now, at least they show what they are.

anyway, if it isn't anthropic it will be someone else, place your bets ...

if you have ever to work with pdf beyond merely staring at one you'll realize to what extent that format is an absolute disgrace. there are a zillion tools out there to manage pdf in a zillion ways and not a single one of them gets pdf parsing and layout right 100% of the time, not even adobe's. the only thing pdf had going for it was that it wasn't msword, and that's why it spread like a virus, but we could really benefit from having a proper truly portable (and universally adopted) document format even at the cost of a reduced functionality or scope.

but the article isn't about that, like the post above yours says llms probably ignore the format altogether and just ocr-parse a printout. what this is about (apart from being clickbait to take you to the verge) is llms having problems in processing text in general exactly the way we would want them to, which is old news. the underlying cause is that llms simply predict text and lack real understanding, which is why their output often looks surprisingly good but is often not quite ok or completely wrong. this is the limit of what their statistical model can offer, now we're trying to furbish them with other supplementing techniques ... with only discrete success so far.

let's say we don't know how to do that ... yet. but we do know about several ways to compromise phones. and we tacitly know there are ways we still don't know. if you have access to the app encryption might not be a problem.

it may not need a backdoor. insider knowledge and some 0-day to get in might go a long way. both exist.

oh, the aim is just to pretend they're fixing the it system, that huuuuge barely tractable problem that eclipses all the endemic corruption in their bureaucracy, government and justice system and their decades long cover-up which was the real cause of all that misery and death. if you can destroy a person's life because a bug in your software, and do it again, and again, and again, and do it over and over for over 30 years then software never was the issue.

but if you fail to fix the system in 5 years, and get another 4 ctos to fail at it again ... after another quarter of a century nobody will remember anymore what the whole thing was about.

it has ofc had some effect on my health, but i'm managing. was/is it worth it? can't say, life is a long winding road. maybe i'm just lucky but i have no strong regrets in that regard, if that helps.

oh, you can hardly become poor from smoking, and being disliked can be just right sometimes! let me show you ...

/puffs into your face

/puffs in your general direction ...

that reality doesn't preclude particular interests in particular segments either, and there are precedents. so a targeted operation on discord is far from proven, but not unplausible.

also (going back to our exchange a few days ago) this instance clearly shows that personal info protection in age verification processes is far from what it is touted to be. this again might be deliberate or just greed and incompetence. in the latter case it would be showing really gross and possibly criminal incompetence, but still, who knows. it certainly is something that raises reasonable suspicions regardless of the motives, if any. (i.e., hardly a conspiracy).

apparently not. half of it seems to be males roughly of military age. i.e., the group you would expect being the meat of an insurgency:
https://thesocialshepherd.com/....

possibly beause he seems less interested in the actual (true) facts than in using them to uplift trump and demean biden (*1). that's a guaranteed reflex trigger for any fan of both mutually opposed but fundamentally equally noxious us oligarchies, and you're guarranteed to find plenty of both but more of the latter here.

*1 given his goal, it's logically irrelevant to him that most of the few tanker seizures under trump had little to do with russia's oil trade, and a lot to do with the currency employed in oil trade in general, and public perception.

btw, eu countries have no way around buying russian energy, nor does it appear that they will in the near future. their only concern is doing it in a way they can deny.