So this customer interaction:

" ... And can you please go die in a fire you complete and total [unclear audio] ... !!!. Thank you!! Next!"

would be scored as very polite by the chatbot.

Here when you buy something you typically pay by putting your bank card into a card reader, or nowadays by tapping the card or your phone for a contactless payment. The card reading terminal of the shopkeeper has has an lcd screen that will show if the payment has been accepted or declined. Does it work differently in India? Do the shoppers have to enter the store's bank account into their phone to pay? In that case the shopkeeper would need some confirmation that the payment has been received.

As much as I dislike defence spending, weapons could at least be useful if Ireland invades, or the Martians, which has similar likelyhood. Or they could sell them to Saudis or somebody similar. Datacenters otoh have zero value, all they do is raise electricity prices. I wouldn't be super surprised if they cause brownouts or water supply problems in this town, UK's ancient utility grid may well struggle to cope with a sudden peak in demand.

Why can't they designate housing "critical national infrastructure"? That's what they should be building in these "gray belt" areas, if they build anything there. This would have an additional benefit of being actually true. Or actual infrastructure, as in roads and railways.

And the usual tortured wording they use

What the fuck does that even mean? How do areas of land "underperform"?

There were some good songs in the 2000's although the decline was already well underway at that point.

Or perhaps "baby don't hurt me, don't hurt me, no more"

That's a pretty contentious claim. You can safely say that new technology has never taken away employment permanently. Or at least long term. The people made unemployed by new technology eventually found new employment, or at least their children or grandchildren did, but short-term automation would often absolutely lead to overall increase in unemployment. And just because permenent reduction in employment has never happened in the past doesn't mean it can't happen in future. There's always a first time for everything.

Besides, the binary employed/unemployed distinction isn't the only important one, the type of job you do is just as, if not more, important. Working some kind of job almost always beats being unemployed, but there is a huge difference between a comfortable respectable middle-class job that can pay for a house, a car and support your whole family and a minimum-wage or below McJob.

The 'increasing productivity' bit is also more complicated than that. How easily you can manufacture a good or provide a service (and usually consequently how cheap that good or service becomes) is only one variable. Another one is the quality of that good or service - and that can obviously be subjective, but often it's pretty objective. Hand-produced nails are no more valuable than machine produced ones (less valuable normally, because human will make mistakes and won't be able to reproduce the required shape perfectly every single time). But with food for example, as you start to add more and more automation and chemicals at some point quality and effect on health can start to go down. Then food produced with less automation can become a premium product.

As far as art is concerned, the desirability difference between human art and mass-produced AI slop is pretty obvious. Besides, even before AI we were already getting way more art than we could possibly consume, and so human art was never that expensive (unless it was collector high art). The phrase 'starving artist' exists for a reason. So increase in productivity doesn't produce that much benefit here, whereas decrease in quality is certainly a huge drawback.

Having said that, 70% of middle class jobs being replaced in last 45 years due to automation does seem like a very dubious claim. It probably depends on how you define a 'middle class job'. Also, if automation replaced one middle class job with another comparable middle class job, would that also count?

I'm pretty sure that wasn't the plan, as far as there has ever been any plan beyond hype the tech and keep getting investor money. I'm guessing they wanted to keep the basic service free to get as many people taking AI for granted as they could, and make their money off premium subscriptions sold to industry. After trying AI though industry quickly figured out that they wouldn't be able to fire all their workers and replace them all with AI agents, as promised, so industry uptake of AI has not been high at all. So now OpenAI and co are probably starting to panic about how they're ever going to repay all the money invested in datacenters, and 'just put some ads in' is a tried and trusted solution to getting some revenue. Even if they planned to start enshittifying eventually, I imagine they didn't want to start so soon.

This is an often repeated but totally flawed argument in that it's trying to estimate AI's 'intelligence' by comparing the text it generates to text a human might generate - if it's similar enough, it must be about as intelligent as that human. If it makes mistakes - well, humans make mistakes too, right? That's exactly what all the firms peddling 'AI' want it to do, imitate human thought convincingly enough, and that is exactly how they want you to think. But we don't need to estimate anything, we know how AI works.

To us, language is a way to express our thoughts and feelings. To us, a word 'dog' is a handle to a cache of memories and mental images involving dogs we've encountered or read/heard about, as well as generalised ideas and beliefs about a concept of a 'dog'. AI does not have such a cache. To AI, 'dog' is an arbitrary linguistic token and it is a handle to a mathematical model of what other arbitrary linguistic tokens go together with the 'dog' token. It's like, some physicist for some reason wants me to give a group of PhD Physics students a lecture about very advanced quantum mechanics, so he types up the lecture and gives it to me to memorize. I do so and read it to students. That means I'm a very knowledgeable in the field of quantum mechanics, right? Wrong - I've just memorized the words without having a clue what they mean.

When people come across false information, they're capable of using their logic and knowledge of the concept that the words stand for to figure out that the information is false and they shouldn't repeat it to others (unless they want to fool others for some reason). Sometimes they fail to figure out that the information is false, because their logic is flawed, because they just don't feel like thinking etc etc. But humans are at least capable of figuring it out. For AI, all information that is in its training dataset is of exactly equal value, it all goes into building the model of how tokens go together. It isn't capable of telling a hallucination from a non-hallucination.

The persons inside/outside the car may be weak, ill or too panicked to break the glass. Especially if the car has reinforced glass windows. It's an action that is a lot more complicated than just opening the door, in a situation where every second may count. The person inside the car may hesitate to break the glass because they're afraid of flying glass shards, or because they don't understand the urgency of the situation and don't want to damage the car any further. Yes that would be extremely stupid but people with that level of stupidity aren't exactly rare.

For example, imagine this situation: an adult driver with a young child (maybe early teens) in the backseat crashed the car. Car lost power and a fire is starting. The adult is passed out in the driver's seat. With manual-only car doors the child can get out of the back seat, open the driver's door from the outside and drag the adult to safety, if that's physically not possible then maybe wake them up. Or a similar situation where a child or an elderly person or someone who would have difficulty or lack the presence of mind to break the glass witnesses a car crash with the driver passed out.

On the flipside, what are you gaining from fully retractable doors? Slightly better aerodynamics? The gain would be very marginal. Better aesthetics maybe? This would be absolutely fine if it didn't potentially compromise the safety of the people inside the car.

Given the quality of writing we've recently been getting he can't be any worse then the randoms they would've given it to otherwise. This way we can at least be sure writing quality will be above fifth grader fanfiction level.

That's a nice idea but I don't believe that's how the c-suite think. It isn't 'we have one revenue stream, we don't need another', nowadays it seems to be 'we have one revenue stream, can we add another? And then another and another?'. See for example paid subscription services like Netflix and Amazon Prime trying to gradually push ads into everything. Free networks like Mastodon might be better.

I miss the old internet, where people would meet in forums and message boards. Nobody (generally) obsessing about likes and followers, nobody trying to monetize every kind of human interaction, just a bunch of people with common interests coming together to discuss said interests. This place, even such as it is, is one of the relics of this bygone age.

Yes that's indeed the goal of every business, but what matters is, how they go about achieving that goal. If businesses try to keep customers by making better and better products, that's a socially beneficial thing, because people will have better and better stuff. If they try to keep their customers by getting them addicted and exploiting their psychological triggers, that's socially negative, because you get psychologically damaged people out of this. You can argue about the extent of the damage done, but it's safe to say that any amount of psychological damage is a bad thing.

As for Meta failing to keep young people, it's arguable to what extent they failed, as far as I'm aware Instagram is still pretty popular. Certainly most if not all young people have moved from actual Facebook to TikTok. That's partly because teens naturally don't want to hang out in the same place (even if it's virtual) as their parents, but I think mainly because TikTok used the same strategy as Meta, but did it better. Better in this context means they banked on even dumber content that assumed even lower attention span and it worked extremely well for them financially. Meta set the standard on how to run social networks/meeting spaces for young people, now anyone trying to create such a platform will copy their methods, except perhaps even more cynically. It doesn't matter if Meta goes out of business tomorrow, the damage has already been done.

They're not building all the data centers for the AI that could solve engineering and scientific problems. Nobody cares much about those except small groups of scientists and engineers. It's all done for LLMs and and maybe a bit for image/video slop generation.

For IT staff, any reasonably organised place will have JIRA or some other project tracking software in place, which will show you exactly who did what and when. Amazon will certainly have something like this. For general office staff there may not be something like this but still a manager who has any degree of competence should be able to say who was assigned what and whether that got done or not (and how well).

This looks like it's copying what DOGE was doing with federal employees last year. I suspect it's preparing justifications for mass layoffs. Unless Amazon genuinely got so bloated and management structure got so convoluted that there are a lot of people who nobody knows what they're supposed to be there for.