It wouldn't be news if you looked at their terms of service -- which you should. The ToS explicitly say they use a combination of automated systems, human review, and reports to identify and investigate violations of their usage terms, including violence, abuse, fraud, impersonation, disinformation, foreign influence campaigns , abusive sexual content, and academic dishonesty. This includes "anonymous" sessions that are saved for a minimum of 30 days. You have no expectation of privacy from the provider's compliance teams.

This is *absolutely* standard among the major online players. So why not use a local AI workstation with a couple of big-ass GPU cards in it to run the campaign? That's what they *should* have done. But the major online players like ChatGPT and Claude are much better at realistic content generation than the widely available local models you can run.

What they should have done is design and run the compaign on a local AI workstation, and used the local workstation to generate prompts they could feed into burner accounts on public services like ChatGPT and Claude. But they got lazy and ran the *whole* operation in ChatGPT, right in plain fiew of the OpenAI compliance teams the ToS they evidently didn't read would have told them were there. They even did *performance reviews* in the same account.

Remember folks, these "spooks" are just mid-level paper-pushers in an opaque communist bureacuracy. You can never discount inertia in such an environment. Because this was something new, they might even have had trouble getting the purchase of some high end GPUs approved.

I don't hate Trump. I pity him. No matter how much shit he slaps his name on, it won't fill the hole is daddy left in him. But people that damaged shouldn't be let anywhere near power.

Well, no. It's true you can't buy books for the purposes of scanning them *and then making them available online* (Hachette v. Internet Archive). Scanning them for AI training is not settled law in every Federal District, although in at least one that has been ruled transformative and therefore allowable (Bartz v Anthropic, Northern District of California).

Anthropic famously bought a lot of copyrighted books and scanned them to ingest into its model training corpus. Arguably they aren't violating copyright because what they are doing is *transformative* -- turning words into a statistical map of word associations.

But what China is doing by inferring the structure of that map doesn't touch on *any* kind of intellectual property of Anthropics. Sure, the map is a trade secret, but they've exposed that trade secret through their public interface. It's not human created so it's not copyrightable. Even if that map were patentable, which it probably isn't, it's not patented.

The worst you can say is that China is violating the service's terms of service, which may have no legal force there.

As a lifelong cyclist, I agree in principle. The problem is over the last seventy-five years we have rearchitected the very geographic fabric of society to make *solving* our transportation problems with bike and public transit impossible.

Before WW2, Dad would leave the apartment and walk or take a trolley to work (usually in the same city neighborhood) while Ma "kept house" -- managed cooking, clearning, childcare, and the family's community and social engagement. In the 1950s and 60s, instead of an apartment, it'd be a suburban house. Ma would drop Pa off at the kiss and ride.

Today Mom and Dad both have jobs they have to get to, usually in *different* suburban employment areas; they can locate to make the commute easy for one, but they keep changing jobs every couple of years while their long-term wealth is being put into a geographically fixed asset: their house. They are financially anchored to their house as their jobs move around the region.

Car-dependency is baked into the very fabric of society, in a way you can't fix with transportation policy or projects--not without decades of projects. But we have reached the limits of the car-dependency model; we can't fix traffic by adding marginal car capacity as has been repeatedly demonstrated by freeway projects that fail to fix traffic because we're in an equillibrium between commute times and job selection.

Transit and bike infrastructure won't fix this, but they *can* make marginal improvements in the traffic situation by taking cars off the road for the minority of people who can use these alternatives at this particular point in their lives. I think e-bikes are going to be key. I personally wouldn't consider a ten mile commute by bike on roads shared with cars a barrier to commuting by bike, but most people wouldn't attempt it. E-bikes on bike infrastructure can make a ten mile commute practical for *normal* people, and take a significant number of cars off congested roads. Public transit could help, but again in a marginal, opportunistic way. In Europe or the US Northeast where car-dependency was overlaid on existing dense urban fabrics, there's a lot of opportunity for major transit projects. But for American cities in the West which have *no* center of mass to build around, solving car-dependency is likely a Moon-shot level project.

Well, that's one hypothesis. However since they saw a significant difference in the population where the social media apps were removed, then if your hypothesis is true, the data would suggest that delivering the service as a native app rather than a web app must have some harmful effect in itself. An alternative hypothesis is that their application usage patterns changed when the apps were removed.

It's not altogether far fetched that web-delivered apps have a different psychological effect than native smartphone apps, because native smartphone apps have greater access to the system for tracking and notifications. Native apps also offer different features than their web versions. This is why I use Facebook via a browser, because the Facebook native app is insufferably intrusive, constantly trying to get your attention. It means, however, I can't use Facebook's chat function.

It would need to be as close to like-for-like as possible or people are going to reject the results, so presumably if they've grown rump steak, say, then they'd need to not just compare it with some actual rump steak, but prepare and cook the two cuts at the same time in the same way. The ideal result here for them here is either "lab grown is better" or "can't tell them apart", after which buying decisions should come down to bang-per-buck, and that might even hold if lab-grown isn't quite as tasty, but is sufficiently cheaper to keep it in consideration.

Sure, they could - and probably will - try and stack the deck in their favour. It is marketing after all. But that can only go so far; if they try and compare a premium lab-grown cut with born-and-bred offal, they're going to get called on that and for many people that will mean that they won't get a second chance, ever. Pepsi was a mostly a pure taste test of two otherwise identical fizzy liquids, but food is really about all of the senses so if they really want to sell this and overcome the ick factor, they'll have precut bite-size pieces of meat ready that look the same, cook them on that stand, and let you compare the appearance, smell, texture, and hear the sizzle while cooking as well as taste the samples.

Supposedly, pork. "Long pig" is the literal translation of the word that was used in many South Pacific cannibal cultures and was used by European explorers to refer to the practice.

It needs a version of the "Pepsi challange" blind taste test from yesteryear. I'd certainly take that if given the opportunity, but have yet to find anywhere with the stuff to try in the first place, let alone to do so in a blind test. If it's equally as good as they claim (and the science says it *is* the same, right down to the cellular level), then they shouldn't have any problem convincing people that it's a viable option to regular farmed meat, and if they can do that, then the cheapest option should win in many cases.

I suspect there may be some legitimate corner cases about "free range", "corn fed", and similar dietry or lifestyle things that will have at least some effect on the texture of the meat (e.g. buff animals vs. couch potato animals), but maybe there are ways to replicate at least some of that in the lab too?

Not yet, anyway, but they do presumably get to see emotional responses like this from humans in their training data. If there are enough human tantrums over code submission rejections in that data, then it's not a huge stretch to where that would be an acceptable sort of content template for an AI to build on to generate a response. A lot of the rest is probably exactly what evanh suggests in their post immediately below; humans using an AI to play games to get a kick out of being mean or, in the case of the ArsTechnica article, a lazy "journalist" not fact checking the quotes from the AI they sent off to do their research and write an article for them.

On Slashdot? Sure. What's more, we'll probably get to have at least one of them twice!

Yeah, some of us used to do that with *NIX systems back in the day. Seperate /sbin and /usr volumes, mounted read-only, and various other volumes, like /home and /tmp, depending on the system use, set to not allow execution. You needed to be root to remount to read-write in order to install patches or updated binaries, then reboot to get back to the read-only mountings. Regular users were not capable of doing jack with the sensitive OS partitions, and most forms of attack were really, really, hard when you couldn't modify any system files or run any scripts/binaries you'd managed to get onto the system.

Used to be you could do that on Linux too, until systemd decided /usr needed to be part of /, which generally also includes a bunch of stuff that has to be read-write. Progress, huh?

Back in the 80's vendors promised companies a "paperless office", but cynics noted that the equipment sold always included printers.

Now we're looking at a promise of a post-human-labor era, delivered by armies of human drones performing mountains fo soul-crushing labor.

I expect that will be a major part of it, but I can see a few areas where AI might be genuinely useful without compromising existing jobs as much as some fear it could. For instance, creation of storyboarding options, replacing modelled and scripted CGI for greenscreen production, creating location shots/B-roll, tweaking existing IRL backdrops to add additional features (e.g. the manipulation of actual shots of places like Dubrovnik in "Game of Thrones", or Valencia in one of the later seasons of "Westworld"), as well as some of the post-production tasks like colour grading and the like.

If they're going to prioritise things like script production (which would almost certainly be deritive because that what AI does) and a reduction in headcount on both sides of the camera (quite likely, given this is Amazon) in order to churn out an even larger pile of mediocre dross for the same total budget then I can easily see that backfiring as the genuinely decent shows could easily end up getting lost in the noise.

The people running these services are essentially re-broadcasting paid-for content for a fraction of the price.

That's quite probably not all they are doing. Remember, if it's too good/cheap to be true, then *you* are probably the product or, in this case, your bandwidth and IP address.