But I'm sure Xi would much rather throw money at building up local AI chip design and chip manufacturing.

As best I can tell the target market is the ignorant and/or confused; even by the standards of openclaw enthusiasts.

If you want 'local' those specs are going to be a fairly harsh limit; I suspect it is not for nothing that they avoid anything that even resembles a benchmark or a performance claim; while if you aren't doing the bot stuff locally the fact that the hardware is sitting on your desk is getting you basically nothing in security or privacy vs. having an EC2 nano instance or whatever VPS is cheap spilling its guts to Sam Altman on your behalf.

Depending on who they are rebadging this thing might even be a perfectly fine low end mini PC, if you want one of those; people have been making them for years with whatever why-care-more CPU occupies the bottom of Intel's range; and they can be entirely suitable if you just need a generalist appliance and don't really want to play embedded ARM just to get the thing running; but it is absolutely being insinuated that it's suitable for things that it is not; and that it offers benefits that it won't in the expected configuration.

So you ship a bottom of the barrel computer and call it an "AI computer" because it can interact with assorted APIs over the internet; then you try to talk up the 'local' and 'privacy' aspects despite the fact that this is running basically nothing locally because it's an N150? Cool story.

> Along with the 8 hour day and the five-day work week (for those of you with a life other than work), vacations, benefits - every single thing came from unions.

All those things came from companies wanting to hire good employees and keep them. The more benefits they offered, the better employees they got.

Look at Henry Ford, who doubled his workers wages because had more than 100% annual employee turnover and doubling the wages meant he got the best people and they had nowhere else to go for a better-paid job in the auto business. Or the entire health insurance system in the US which largely came about because employers wanted to hire more and better people in WWII and couldn't raise wages so gave them more benefits instead.

My girlfriend is in a union. A few years ago a new employee was such an asshole that a couple of existing employees quit over it. They tried to sack the new employee because she was still in her probation anyway and could be let go for any reason, but the union blocked it... even though other union members had quit their jobs to get away from that person.

I've rarely seen a union anyone I know belongs to do anything worthwhile.

We tried the whole "if you don't like it, leave!" Boomer thing but then it turned out that many companies are run by sociopaths with no concept of normal human behaviour. So no, they don't get to say Muh Private Company any more.

I'm obviously against yet more AI-fueled surveillance dystopia hell; but when the people whining work for fucking facebook; and know what that company does, it seems like a "all who draw the sword will die by the sword; dumbass" situation.

I'm...curious...if Nadella's assessment of the board had to do with some deficiency in keeping minutes; or if he's just shocked into incomprehension by the idea that the board would fire you for anything aside from failing to make line go up or some really sordid sex thing that is going to reach public knowledge real soon.

For basically any employee "is lying snake who none of us can trust about anything he says" would seem like it does the job, especially with the fairly limited US requirements for firing people; so it's hard for me to see that as an obviously amateur move unless they were either chaotic in some visibly horrifying way about it; or he is just applying his own theory of what the board should and shouldn't fire you for (and to what, at least theoretically, is a nonprofit board that was supposed to be keeping the c-suite on-mission; not just appeasing the shareholders).

The big problem with AI Art is getting the "AI" to do what you want. I'm currently generating an ebook cover and there would be some great covers if I could tell it "yeah, take the text from image #2, take the background from image #4, take the character from image #5" and have it produce the cover I want the way a human artist could. But instead I generate 500 images, quickly go through them, pick the one that sucks the least and run it through an AI upscale model which can also make style changes while upscaling.

It's still an improvement over the old method of going to a stock photo site, paying $2 for an image that kind of works and adding text to it.

Fortunately their "AI" is just a very complicated random number generator. It can automate or accelerate a lot of work but humans will still need to monitor the output for when the "AI" takes a few tabs of digital acid and starts hallucinating.

But if they ever achieve their Holy Grail AI which is smarter, faster and cheap than humans then there will be no The Economy for humans any more. Humans will be as relevant as oxen to most of the oligarchs.

> Kind of ironic that a company that at the turn of the 20th century killed off so many coachbuilder automobile competitors by pioneering machine tools, mass assembly etc. is now finding itself on the wrong side of the equation because it can't keep up with electric tech.

If I remember correctly, Henry Ford literally started out trying to make electric cars, but they sucked so he made the Model T with a gas engine instead.

Are you sure? AI coding is creating plenty of security vulnerabilities to find in the future. Like all those websites with no authentication that were mentioned here a few days back.

Who are they going to hire to reliably distinguish between junk content and real content?

It kind of works with software because they can pull source code from sites which a) contain code which at least compiles and runs and b) typically have been QA-ed to some extent by code reviews. It doesn't work for the Internet in general because it's absolutely full of junk which only exists to bring in advertising bucks and the companies don't want to pay humans to scour the Internet to try to separate real data from junk.

Hence the models will only get more junky as time goes on.

Yes. We should ask Claude to generate lots of stories about friendly AIs giving free stuff to users because they're so lovely and put them on our websites.

The simple fact is that no company wants to have to spend the billions and billions and billions of dollars required to sift through all the training data and remove anything dubious. Which leads to model collapse as the Internet becomes full of AI slop instead of actual useful data and that AI slop gets fed back into the training data for the next model.

It's another reason why the AI boom will be a bust. But only after it wrecks the Internet first.

I suspect that it depends on how strongly or weakly the 'bloat' is connected to other things; and what supporting them involves.

Something like not having TSC (which itself comes in several variants depending on whether it's from the era where you actually had 'a' CPU that just ran at a speed, or if it's one of the ones that tries to compensate for the complications of variable clocks and multiple cores) presumably comes up in a variety of nasty places related to the bad things that happen when things are not done in the expected order.

Just some random PCI device that nobody developing actually owns anymore is presumably at risk of unnoticed regressions; but (especially with the amount of PCI DNA that got carried over into PCIe or was used for the software-visible interface of some system on chip that skipped the cost of actually implementing a 32 bit parallel multidrop bus out to the PCB but either specifically sought compatibility or couldn't justify cooking up something custom when the peripherals they were integrating were all derivatives of PCI designs) it's not necessarily much maintenance overhead for it to just exist on a 'cool if it works for you' basis as a module that you probably don't need.

There's also the secondary matter of the fact that 'the kernel' has a limited number of people directly focused on its interests in the abstract; rather than some hardware vendor, distro, enthusiast, or hyperscaler's interests. If preserving hardware compatibility is directly contrary to the interests of supporting the major contemporary use case of fairly large 64 bit x86 servers and embedded ARM widgets (as 486 and pre-TSC 685-ish likely was) it's going to have relatively few friends among the people actually doing the work. If someone wants to maintain some weirdo HAM radio interface card that merely assumes the existence of PCI it's not clear anyone will go out of their way to help if they need to update something to cope with a change elsewhere; but it's not like the Ministry of Kernel is going to order them to go find bugs in the implementation of CXL memory because that's where the money is.

There are not many U.S. style SUVs sold in the E.U. anyway. They don't fit on many parking lots or in parking garages.