Replaceable? No. Reparable? Depends on the extent, but even that's hard: the wings are full of hardware, and if you have to spend a year and invent a process for dismantling everything to get at the damage, it becomes financial infeasible. Even if you pull it off, you have new inspection requirements, operational limitations, etc.: it's not the same revenue generating plane after something like this.

There is a lot at stake. Emirates operates these with over 500 passengers. If that manifest burns to death on takeoff because a wing folds there will be hell to pay.

AI is stupid, but it can track details and at least provide a hint.

On the "make stuff that works" side, I've had the experience like yours. If someone lets it go too far, then it's a headache and it's easier to roll it back. If anything a bit painful as it's "empowered" some of the worst to make my day harder, being confidently incorrect all the time.

On the "find stuff that's broken" side, well, humans don't have the attention span and the AI techniques are catching little but critical mistakes by humans. In my codebase that has had regular security reviews for over a decade, including a few outside consulting companies, this year marked the first year that the teams had LLM at their disposal. It found two security issues that no one had noticed, one of which had been there almost essentially from the beginning. Admittedly, neither were exactly world ending (both required attacker to log in with admin privileges and the things you could do were a bit constrained), but they were real and undesirable. One of them kind of missed the point and although it mis-characterized the behavior, it put a light on an area where a human could actually think through and sort out the real issue, and how the flawed approach applied more broadly than the AI identified.

So AI as a review tool can work, it has a significant amount of false positives and misses stuff of course, but it can either directly catch or inspire human attention on a sketchy area.

The security team said that it was actually quite remarkable they only found two vulnerabilities that both required admin access, that most projects they dealt with ended up with 4-5 vulnerabilities exposed to either unprivileged users or even unauthenticated access. This is the sort of code the world is mostly built on.

Probably heard it in church Sunday School.

Conversely, if a Mac costs you 1 hour per week because it doesn't do something you need it to it costs $5200 in productivity and the cost of the Mac.

And it does cost in productivity as one of my illustrious duties has been to maintain a Windows Server RDS farm solely for Mac users to be able to log on and use the same applications as everyone else in the business. So not only was it lost productivity, it also cost my time (which is worth more than theirs, I get billed at £180 per hour) and a Windows RDS license. All this because the special little sales people didn't want to use a Dell like everyone else.

Apple, the product that keeps on costing.

Yep, this is blatant protectionism, which is the 'Murican way. This is why Japanese diesel pickups have been banned using a variety of increasingly complex and convoluted rules designed to not prevent US pickups from being banned under the same rule. This is done because the Toyota Hilux would decimate Ford F-series sales, not just immediately but over the long term because you don't have to replace a Toyota Hilux every year as they don't fall apart.

National security is the sledgehammer approach. They're not even bothering with convoluted rules so they can pretend it's not blatant protectionism.

If anything, the data of American drivers is safer in Sweden as they're subject to EU laws like GDPR. I'm certain Ford, GM and Tesla are selling that data to all and sundry.

Sounds like a market opportunity!

"the greatest arrogance of all: “Save the planet!” What?! Are these fucking people kidding me?! Save the planet? We don’t even know how to take care of ourselves yet! . . . The planet will be here, we’ll be long gone; just another failed mutation; just another closed-end biological mistake; an evolutionary cul-de-sac. The planet will shake us off like a bad case of fleas, a surface nuisance." - George Carlin

JIT emulation is also interpreted code. Not everything that is interpreted code is something high level, like javascript or python.

As for apple's relaxation of interpreted code, they still put the kabosh on most of the more useful applications of it.

Like JIT.
https://oatmealdome.me/blog/wh...

which is why I can still do more with my ancient pixel 4A, than you kids can do with your latest edition ipad/phone.

Apple's walled garden comes at a price.

I'll definitely give you these ones.

PDF manipulation on Android is *TRASH*, and windows 11 is a horrible operating system, that wants very much to treat you very poorly.

Sadly, device makers really *REALLY* do not want to open up their tablets for Linux like they should. Touch interface on linux might be a bit rough, but there are much better tools for office productivity and PDF manipulation for linux, than for either windows or android.

I just dont feel that I would be willing to fork out the extra dough for an apple tablet, where I would lose all manner of other functionality, just for PDF and office tasks.

While support is limited to a handful of models, due to a shortage of skilled maintainers, there are a few android tablets that support lineage OS, or postmarket OS.

Like OnePlus tablets, and Galaxy series tablets from samsung.
Good support from LineageOS for those. The Galaxy Tab A7, from 6 years ago, *IS STILL GETTING UPDATES* from Lineage, for instance.

LineageOS on android devices push their service lives *WAAAY* past what is normally doable. My now very ancient Pixel 4A is *STILL* getting monthly updates via LineageOS, for instance.

"nothing gets fixed"
Say again? What part of Monthly Updates did you not understand?

*IPv6 support*
My phone supports it natively. Dunno what you are going on about.

*superior AI*
This is a matter of personal preference, and I will assert that my preference trumps yours, when it comes to the device that *I* am using. Thank you very, very much.

*Able to play older games*
I can sideload on old android 2.x games on my phone fine-- Oh, wait, I said sideload, Thats a thing apple products cant do, isn't it? Awww.

*Unless they are old mac games, ironically*
Must really suck that Apple wont let you run interpreted code, huh? As for myself, I can run all kinds of stuff. Emulated game consoles, Winlator for older windows titles, BasiliskII for that classic mac experience you mentioned, ADosBox for classic DOS titles, all manner of stuff. Real shame apple doesn't let *ANY* of that play on their products!

*Able to use UNIX tools because its UNIX*
You do know that you can in fact get access to the console, without rooting, on android devices, right? There's any number of solutions for that, of varying quality, from the playstore-- and if you dont like any of those, you can sideload (Ohh--- there it is again!) F-DROID and get direct-compiled FOSS tools.

*Superior Hardware*
In what ways, exactly? They are all sealed units that you cant easily open, All have unremovable batteries, etc. If you mean "It has an alumalloy frame", I hate to burst your bubble, but there are plenty of such tablets in the android offering space. A great many android devices are built around very mature 3D rendering capable graphics chips as well, so when you load up things like Winlator, you can actually do D3D games on them. Not terribly well, since its ARM64 emulating an X86-64 CPU, but good enough for most things that you would actually find pleasurable to play on a tablet form factor.

*High performance graphically and otherwise for the cost*
Look above in the comments for how a savvy buyer saved over 1000$ buying an android tablet, then get back to me on that.

It's UNIX! I know this! (sound of crashes in the background)

It's just preparing a narrative in the media. No different from what Anthropic did with Mythos and Fable. "Waaa, waa, we could so make money if the bad competitors didn't thwart us and bring our profits down by luring our customers with lower prices"

Remember: it's IPO season for Anthropic and OpenAI. Anything they say now is (more) suspect (than usual).

I think the end state of all this looks like game cartridges. If people could buy the weights of a frontier quality model in a high density, high speed ROM that ran locally plugged into an PCI-E or M.2 slot, all you would need then is a reasonably fast tensor processor with a little RAM for context. This is possible, and there is even a company (Taalas) with an early product. They have an online demo that is crazy fast.

You would buy one and use it for some time, probably years, and then discard it when the value of some newer version makes sense to you.

Not far from the truth.

The big implication from this is that making these inference engines isn't difficult and they can't be protected. That's the doom of all of AI money: the product is getting commoditized; you'll buy a Claude/ChatGPT/whatever in a box from Amazon in a few years.

Presumably, this is the whole point of these contracts, to hold the customers accountable for Micron making memory for them when the broader market is not necessarily looking for that particular memory.

Now I don't see how this can work out in one of the more well documented ones. OpenAI had at least a trillion dollars of these sorts of purchasing commitments, and even pretty bullish assessments don't support their ability to make that much purchase. So I do anticipate the market failing to make their minimum purchase commitments. I'm presuming there are penalties in these agreements and so they probably get money for nothing unless it gets so bad that OpenAI goes actually bankrupt.