Another small point... if you want to put something in the Apple store, you'll need at least a small Apple device to be able to compile it before you submit it.

I suspect you may need to still do this if you were to submit something to the AnonymousCowardStore too - although at least there it would be possible for a third party compiler toolchain to do the work (how possible? I have no idea, but at least the store wouldn't reject you out of hand for trying).

> so the cost is socialised to the British taxpayer.

The tax payer is paying *something* here, but it was a loan, so financially the tax payer won't be losing out. You can argue about the opportunity cost of that 1.5bn, and that's definitely valid, but the 1.5bn should be coming back to the people.

As noted elsewhere, the loans were to JLR suppliers, not JLR directly (because yes, Tata had to do that themselves). There's a lot wrong with a lot of this story, but on this point, I find it hard to really criticise the actions of the government (as far as they went, at least). I have no idea if they're able to do such things, but it'd be great if they could now hold JLR to the coals a bit to ensure they get all those suppliers paid so they can pay back their loans super quick (and maybe as a deterrent for next time?).

It could also be a genius move. It's playing to billionaires egos - their dick measuring can now be put to good use getting humanity back onto the moon. It does feel like no one else can really compete here, but ego seems to be a powerful motivator, and at least one rocket even looks like a dick. You may be surprised what comes of this.

Then there's the conspiracy theory... Trump and Musk fall out, Trump makes a call to Nasa... contract re-opens. Either way, same outcome.

I suspect a lot of times, in peoples minds, "capitalism" is wrapped up with "democracy". They're not actually dependent on each other, although they do go well together.

The rich may well just be fed up with democracy - they want to do the controlling, and don't want a change in leadership every few years which may more may not suit them/do what they want. Thus they're looking to weaken democracy so that votes either cease altogether, or become so meaningless that the majority don't bother to vote. They'll then pressure the leadership to promote them over others, which isn't capitalism, but it's not socialism/communism either - it's perhaps just abuse of power.

If only they'd hired you to design the main computer for the sub, eh? We might have been watching the 'cockpit recorder' and analysing telemetry instead of nothing much.

Fewer GPUs = lower power consumption, for sure, although of course, more of the transistors in those remaining GPUs are active now. Either way, the difference in quantity is so great, there *must* be a power saving here. The cost of keeping them switched on is appreciable, and I assume you can only put a certain number of GPUs on a given motherboard, so you presumably can have less servers running the GPUs as well.

Either way, this sort of thing can only be a good thing for the world because of reduced consumption. Looking forward to seeing them in use everywhere.

> ...and creating complex Excel formulas

By christ, they should put 'guard rails' in that stop you ever doing anything complex in Excel. I dread to think of how much of that bank relies on cronky spreadsheets instead of proper, auditable, tested systems.

You: Create an excel formula to take X and Y and Z, and amortise over 6 months, then leverage A, B and C and exponentially decay adding those month on month to what you first came up with.
Copilot: Certainly! But first, please give me the magic number from your manager authorising this insanely stupid idea

Indeed, which is why we should definitely modify humans to avoid climate change. The modification isn't even invasive, can be performed remotely and has some potential to be self-replicating. All you need to do is to modify an ignorant human into an enlightened one.

Climate change is happening, and no amount of "it's a con" or other deflection is going to stop it. We know that humans are at least in part responsible for it, we know what sorts of problems its going to cause us, we also know the primary mechanisms that are causing it, and so we should all be working to solve those problems as quickly as possible.

The best bit is we can all work on an area we like the look of. We can work on reduction of emissions, or sequestering, or energy efficiency, or just simply reusing and repairing old stuff instead of buying new. What we can't do though is to keep emitting as much carbon as we have been.

At least one of their competitors already does the same thing. Also a 'premium' product, with crappy support, now papered over with AI.

> They're just advertising that their customer service sucks

Or that their product is too complex or not intuitive, causing support calls for stuff users should be able to work out for themselves. Failing that, their documentation is rubbish, or quite probably (knowing how it is with a lot of companies), the search engine on their support website is a crock of shit.

Ultimately, at least 80% of the support now done by AI should either have never needed support, or should have been solved by documentation. The last 20% is for the idiot users who are either too lazy or can't work the search engine, or seem never to have used a computer before - there I can see AI being rather helpful.

As an aside: I've been working with one of their competitors - their support is also terrible (even the AI suggestions), but their search engine is absolutely the worst kind of trash you can imagine. They have documents for just about everything, but of the last 10 support calls I've logged, probably 7-8 of them have been solved by the call agent finding a document that I couldn't (and in a few cases answering a question that wasn't addressed in the documentation). These are some of the biggest tech companies in the world, and they don't seem to get this stuff even vaguely right.

I have no idea how this pans out, but I'd imagine their first phone won't be a smart phone. They've got far too many 'low level' issues to work out first, and even getting hardware which doesn't need a manufacturer's binary blob to make it work is already a significant challenge. The subject of apps, app stores and operating systems will likely come later.

As others have said, the low level radios and modems have the capacity to be a real problem for networks. Most networks are held together with sticky tape and chewing gum - the only thing keeping them working is that all phones "play nice". They're not under any obligation to let "unknown" devices work on their networks, and so it remains to be seen how the FSF will make something "open", yet "controlled" enough that it's not easy for someone to clobber a mobile network if they start fiddling with it.

If I had to guess, I'd say the FSF will find a suitable hardware solution, which needs a signed binary driver to run. They'll then open source the binary blob, but a small group of developers will be able to kick off builds which sign blobs. The rest of us will download an FSF blob - we won't have access to any of the internals of it. But that's all just me guessing - I'll look forward to seeing how they solve these sorts of problems.

Indeed, getting line of sight when there's a 50 storey obstruction next to you is going to be tricky. So is signal bounce. GPS receivers don't really have the concept of "signal strength". That is, the GPS signal is below the noise floor anyway, so you can't really measure the strength of it. Instead, you have to spread-spectrum hunt for the edges of the digital signal, and when you've got a few edges in a row which look like they match the pattern you're expecting, then you declare that signal "locked on" and can start to work out the time difference between it and other signals.

So yeah, doing any sort of traditional signal processing to try and work out what are direct and what are bounces... that's tricky. More satellites is handy, but if you're in a tall building corridor, then it's likely you just can't get the "other two points of a triangle" to be far enough away to make a proper location calculation. Mixing with other data sources is probably the answer - cell towers would seem a good option, as would wifi and even bluetooth signals (which is one of the ways Google Maps can do it, because they have maps of wifi network names). Indeed, most map solutions use the fact you're moving to avoid putting you inside a building, or whatever too, so they're trying to make intelligent guesses as to where you are based on the context you're in.

My Audi had a purchase option for an additional speaker system to play engine sounds (which are only played inside the car - they replace the back seat armrest with speakers, and put a couple more under the seats, IIRC). It was something like an extra £400-500. I mean, are you like 8 years old? You need your car to go "brum brum" and make squealy noises when it goes around corners? I'd be *embarrassed* to have such a feature, hell, I'd probably pay to have it removed if it came with it. Besides, one of the wonderful things about EVs is that they make so little noise.

Audi also want to charge me to change the background picture on the infotainment unit too. They clearly have very little idea what sort of things I'm likely to spend my money on. I doubt Ferrari ever really need to worry about my tastes, but they also seem not to know much about people like me.

I wonder if they'd be *more* successful with shorter copyright?

If copyright was (say) 20 years, then fans (and AI companies!) could be making all sorts of derivative works from 20 years old stuff, instead of stuff made more recently. It's probably possible for the AI scrapers to roughly even get those dates right themselves - if it looks like $film_released_before_2005 then it's okay, otherwise no. As it is, copyright is so unimaginably long, that people are just ignoring it completely, not least because working out when something may or may not drop out of copyright is too hard, and of course leaves you with such a small training set as to be pointless.

Of course, they're likely far too bone-headed stubborn to even consider such an idea, let alone go along with it. I'd bet they'd prefer to lengthen copyright to "keep AI out of it" before they'd ever think about any other options.

The more niche the subject, the easier it is. That is, if you try to poison the LLM on the works of Taylor Swift, it's likely to be much harder than on (say) the way to kill a process on an operating system. In the TS case, you've got to overcome an awful lot of very popular 'real' content for the LLM to consider your slop to be a likely extension of whatever it's spewing out. In the case of killing processes, there's a lot less of that information around, and so poisoning it is (relatively speaking) easier.

I wonder how long it'll be before the troll farms are able to manipulate AI on 'big' subjects like world leaders, or government records on a variety of hot topics?