I'm curious(honestly a bit morbidly) to see whether there will be organizational/emergent downsides. In principle not being able to do novelty is fine. Realistically a lot of science simply isn't going to be novel; between interesting hypotheses that ended up not being borne out by experiment(such 'negative results' being a necessary but under-published area) and things like ecology or material science where sometimes you just need to characterize all the bivalve species on a coastline even if they seem to mostly just be fairly basic clams and mussels; or pound through eleventy-zillion variations on plausible engineering ceramics looking for atypically good properties.

However, as we've had the misfortune to see in our inboxes and the powerpoint decks that get inflicted on meetings; people tend to generate more of what is easy to generate; and scientists often get rewarded(in terms of hiring/tenure track; grant funding, etc.) for apparent productivity. The most extreme exploitation of 'AI' for this has proven somewhat nontrivial to combat(the various shit-tier journals and paper mills that add noise to the literature and allow the dishonest to pad their resumes existed back when they relied on human labor; but systems that are good at high speed production of plausible-looking output have thrown them into overdrive); but even if you focus entirely on honest actors doing real science in good faith, if the bots are genuinely useful but only for non-novel work; that seems like a situation where you've just created an incentive toward focusing on low-novelty work and toward seeing the scientists doing the most bot-friendly stuff as the most productive; while those who venture beyond the scope of its abilities are putting out papers by hand and appear much less productive.

Science has often had a slightly ambivalent relationship with novelty(you probably won't end up being remembered as a rock god unless you do come up with something really cool; but initial reception can be downright chilly: look at the career trajectory of someone like Barbara McClintock and transposons. Worked out for her in the end, Nobel in 1983; but that was for work that people were actively disinterested in when she did it in the mid 40s to 50s; and not everyone gets brought in from the cold before the point is moot); but it's probably not going to help if, wholly aside from cranky old guys controlling tenure committees and grant allocation, people who hew quite closely to data that LLMs have already chewed over genuinely crank out research substantially faster than people who venture further afield.

I'm genuinely curious if there will actually be a reallocation to 'AI-focused roles and initiatives'; or if that just sounds nicer than "business is bad and we've run out of room to pretend otherwise"(or the very similar "business is bad and we've run out of room; but we are using a transcription bot in some zoom meetings" which is an 'AI initiative' in the sort of exceptionally banal way that farming some stuff out to AWS is exciting cloud-native hyperscaling)

Aside from skepticism founded on the largely negative results reported for business 'AI initiatives'; I'd be very, very, curious whether there's something cool or profitable you can AI out with pinterest internal data(I assume that their tracking of user interaction and click-through behavior and such is more precise and detailed than 3rd party tracking); but can't with a nearly-free scrape of the parts of pinterest that are publicly exposed on the internet.

It's entirely possible that Pinterest is single-handedly responsible for most of what AI models 'know' about how mormon housewives classify conceptually related images; but even if that is value it's not something pinterest is going to capture; it's just going to get scraped out of them.

I can't comment on whether it is 'desperation' or not; but this seems very much in line with how they handle the "smart features" toggle in 'workspace'/gmail. There is just the one(merrily enabled by default in the US; not in the EU) and it gates both things that fairly obviously involve their bot like "Smart Compose is personalized to your writing style" and ones that we somehow managed to implement a zillion years ago by banging floppy disks together in the dirt "spelling"; along with ones that pre-'AI' google somehow managed by other means, like regexing tracking numbers out of shipment notifications.

Hasn't quite gotten to the point of locking support for signatures behind 'smart reply'; but the structure of the consent-to-all-or-get-none checkbox is certainly quite opinionated UI design.

Any word on whether the implementation is actually invasive, unreliable, and unbelievably computationally expensive; or is a desperately classical feature that's probably older than I am being artificially tied to a totally unrelated set of antifeatures to help goose adoption numbers?

Only if you are willing to accept specific drawbacks.

If you are just doing it to save bandwidth the concept of shipping encrypted data in bulk and delivering a tiny private key over the network is viable; but that does mean making the physical release literally useless without a network-connected install (potentially a full size one, depending on whether your encryption interferes with layout/seek optimization on the disk and what the console's support for partial disk/partial install setups is); which is a potential issue for some physical copy buyers. It's not news that the release version normally gets a more or less massive patch covering all the changes made while it was being mastered and duplicated; but when someone is making a partially emotional decision between physical and electronic copies the difference between 'v1.0 is scuffed' and 'disk is literally unplayable without install and internet connection' can be significant.

'timelock', specifically, also addresses a questionably useful niche: there are cases where the ability to keep a secret for an extended but defined period of time is useful; but if you are a game publisher developing a game with hundreds(probably thousands if you count various support studios and support staff) of people across multiple sites it would be an...atypical...IT situation if you were somehow buttoned up enough to avoid leaks during development; but not buttoned up enough to hang on to a single private key that needs to be used once and kept secret for less than a month. It would, presumably, work; but it's not a situation where the security assurances really line up with the threat model all that well.

> ubisoft is in fact not in your house with a gun to you head demanding you play assasins creed 15 but if someone else enjoys the "woke nonsense" who are you to complain, just dont play it, let other people enjoy things, nothing is ruined for you.

Isn't that exactly what happened?

Setting aside the absolute purist position of Stallman, the point of the copyright law was to give protections to the creators with the understanding that the created works would enter the public domain so that they can constitute a common culture. This was initially 14 years, which was quite reasonable, that later turned into life + 70 years, that is the cause of distrust in that law.

If you polled non-teenagers, I doubt that the majority would suggest a 100 year copyright protection was a fair period. Laws should reflect the consensus of fairness, not behind the scenes deals with politicians by organized power centers.

> gamers suck, the worst and most annoying and most entitled group of fans who never know what they want and demand everything

The customer is always right.

It's not a monolithic group. Almost 50% belong to it. The gaming marketplace has a lot of choice. So studios actually have to compete. They should be glad that their customers are letting them know exactly what they want and if they can't, they will lose business to someone who can. That's the essence of a free market, no?

> all they demand is whatever the next trash from unisoft has

They are not, which is why Ubisoft is crashing to the ground.

> while not paying any attention to the entire mid and indie level scene where there are also really good movies being made

But that is exactly what is happening. Indie segment has had a resurgence because AAA has lost the plot. Indie segment is set to double over the next 5 years while Ubisoft is going underground. That sounds exactly like what you want.

"We believe that the moral status of AI models is a serious question worth considering."

It's honestly hilarious to watch the AI guys talk about 'sentience'; because they are in the simultaneous position of trying to talk up how smart their product is and trying not to say "slaves-as-a-service" out loud. I'd assume that, were customers to actually become confident in 'AI' tools' ability to not fuck it up without constant supervision that frequently ends in scrapping it and doing it yourself; it'd be a full court press to convince people that, obviously, brute algorithms couldn't possibly be sentient and the very idea is preposterous.

Because China values North Korea as a buffer state between it and South Korea. And China is a veto-wielding permanent member of the UN Security Council.

East Germany unified with West Germany because of the Soviet Union's acquiescence. And the USSR, now Russia, is also a veto-wielding permanent member of the UN Security Council. All 5 veto-wielding permanent members of the UNSC must agree before any planet-scale action is taken.

I suspect that they don't really care. There are ways to add a little, um, 'tilt' to actual law enforcement(perhaps the most common involving the case where the number of violations exceeds the time available to process them all; so you can do entirely legitimate prosecutions and still be selective); but it's ultimately limited by the fact that you are forced to refrain from some of the more overt and egregious unprofessionalism where the judges can see you; and people can avoid your reach by being genuinely not guilty.

This creates the temptation to simply not bother; and do as much as you can get away with to make mere contact with you the punishment. Sure, that hoses your odds of getting a proper conviction, you'll be having evidence struck and dealing with arguments about juror partiality forever; but that's the point: you don't care because you are just interested in harassing whoever it is you dislike or who doesn't kowtow properly; and nobody can beat the ride, even by the most ironclad innocence.

If you are going to be a proper goon squad; you are probably downright offended by the idea that issues with securing convictions would be a real problem: that implies that there's still a legal standard that is theoretically the highest authority; with the judiciary working from that; and you are just handling implementation, like a clerk. Simply ignoring prejudicial effects on the case allows you to make the exercise purely one of force against your enemies; which is so much more satisfying; especially if you think the leopard will never come for your face; and impunity will always remain at your back.

"Korea's approach differs from the EU by defining "high-performance AI" using technical thresholds like cumulative training compute, rather than regulating based on how AI is used. As a result, Korea believes no current models meet the bar for regulation"

Call me a cynic; but I'm not sure I believe that people running even-bigger-than-the-ones-that-cost-not-necessarily-single-digit-billions-per-year models will be unduly perturbed by the possibility of maybe getting a $20k fine at some point in the coming years. That's not even "ah, legal for a price" tier fining; that's some sort of weird deadpan comedy thing.

Given the amount of benchmark-gaming and the opacity of the big name hosted operations it's a little tricky to make precise comparisons; but the bigger question seems like "will it matter?"

If anyone still thinks that LLMs are just suddenly going to Singularity and turn into an archotech omnibrain then, yes, it matters a lot who crosses that line first(or it might; also possible that cyber-god will be totally uninterested in who its nominal owners are and have something else in mind; unknowable alien superintelligences are guesswork like that); but if the trajectory is a sort of muddled meandering toward sucking incrementally less; with it being discovered that lazy drop-in integrations actually work really badly and nontrivial fiddling is needed; it seems like a market where the expensive, controlling, probably-stealing-your-data-for-training-or-to-replace-you first movers are going to be less attractive than the runners up just because those guys are hungrier and more willing to give up control in exchange for money.

Would I rather have the 'best' model, as provided by the guy who is just pinkie-swearing that he's not mining my every interaction with the bot to replace me the way Amazon uses 3rd party sellers to determine what categories to slap a house brand competitor on; or a somewhat less good one from some dude who is hungry enough that he'll give or sell me the weights and let me run it locally or on infrastructure that's just someone else's tech swapping dead GPUs, rather than a hyperscaler reading anything they want?

I assume that the plan is unnecessarily stupid and involves 'AI' somewhere; but, especially for de-facto-standard software whose customers find it pretty much mandatory it seems like there's potentially a lot of slack in the sales process.

Maybe it makes more sense in cases where 'sales' includes a fair amount of sales engineer to help customers who don't know the details of specing a full virtualization infrastructure or something; or cases where which SIEM the customer picks will involve some careful wining and dining of C-levels; but I'm perpetually amazed at how much faff there is in 'business/enterprise' procurement vs. smaller scale. You want 10 laptops? Newegg wants your credit card number. OK. You want a hundred laptops; or 10 laptops from SHI or Ingram Micro? Well, the catalog is full of lies(prices, availability, or both) so talk to your rep: then receive some quotes for a variety of things tangentially related to what you need; then clarify that "I need some T-series thinkpads" does not mean "tell me about lenovo chromebooks"; then take a look at the second round of quotes; then maybe cut a PO for one of those; or maybe it's near the end of somebody's quota cycle and someone wants to haggle. It's exhausting and an amazing waste of time.

I assume that there's some utopian/dystopian dream of perfect price discrimination through constant high-touch handholding; but it all seems so futile when the 90% of customers who are less influential can probably just be handed a price and told to like it; maybe a couple of bundle options if you want to get fancy; and the 10% of biggest/highest-clout ones are mostly just going to get some reduction off list price or shell gaming of how much a multi-year contract costs on each year to help them smooth over their particular purchase approval process; and yet you have legions of FTEs who get paid at least lower white collar wages(sometimes quite handsomely) to overcomplicate things.

Early adopters of computers did not get much of a return of investment either: expensive machines, each vendor had their own ecosystem, far fewer features than today, employees had no prior skills.