In theory I think it was more energy efficient, giving them a very slightly longer battery life. Plus there were probably supply chain reasons for it too, such as allowing them to stop making the older chip while continuing to make the Vision Pro.

The Vision Pro has always struck me as a device in search of a purpose. I think Apple was hoping someone else would figure out what it was useful for an then swoop in and Sherlock them, but so far, no one really has.

It is a vast majority. The midterms won't go like people on the left expect. There's one group that's hated more than just about anyone else in polls of Americans: Democrats. They manage to be less popular than Trump and less popular than Republicans.

People may not be terribly happy with Trump and the way things are currently going, but it won't take too much to remind them how much worse things were when Democrats were in control.

Despite his current negative approval rating, Trump manages to be one of the most popular politicians in America right now, even with a net negative approval rating. That mostly because Americans just do not like their current politicians than Trump, but ultimately, there's a reason Trump won in a landslide. Americans may not really like Trump, but they loathe the alternatives.

I have a need for thousands of VMs for a load test in Azure. But Azure requires dual-stack for IPv6 to work -- which completely defeats the purpose! The DevOps team tells me we are out of IPv4 addresses. If everything IPv6 also requires an IPv4 address, then IPv6 is useless. I am told that AWS VMs do not have this problem.

This is a valid retort. But let us not think that lawyers are struggling: once they get to be a "partner" in a firm they are likely making $1 million/year. And the entire context of the discussion is that they aren't relying on staff like they used to. Back in 1980, a lawyer had staff members who ran down to the court house to get documents, bring them back, photocopy them, staple them, file them, make phone calls. Now all of that is 100% automated, plus now they have AI.

I'm not sure the legal overhead is quite what it used to be.

Disclaimers like this apply to Excel, TurboTax, GCC, ChatGPT, and more: The user is ultimately responsible for the application. The manufacturers always disclaim responsibility.

You can get companies to stand behind products and accept liability or sign a Business Associate Agreement - but you are going to have to put it in a contract and pay extra for it. This is why the product you buy at Home Depot and the one the government/military/NASA buys has a very big price difference even if it is the exact same part.

Every industry does this.

From Housing inspectors and plumbers, to software products - it is super common. I just had plumber put this into their contract for replacing a cast-iron drain with PVC. Then I had the tub reglazed and they did the same thing. There are often two prices, based on if you want a guarantee behind it or not. I paid a structural engineer to inspect the foundation of my prior to purchase. While he said the cracks were normal setting, the price was $200 for the inspection + verbal assessment, or $600 to put it in writing and stand behind it. In the last two weeks I've gotten this same thing from a tax preparer and a property attorney. Free advice from the tax preparer, but if we want him to file it and sign it there was a price. The attorney told me what to say in court, but quoted me a price to put it on his letterhead or to show up and say it.

1. Raw work: 8 hours
2. Work with unchecked AI: 8 minutes
3. Work with check AI: 16 minutes

I don't get why people choose option 2 over option 3.

They stocker making $15/hour needs to work extra hours to survive. Why does the lawyer making $500/hour overwork?

Not to mention that the first thing any gamer does when they get a game is turn all that artistic crap off, both to get a better framerate, but also to make the game easier to see. The fewer "artistic effects" on the screen, the easier it is to see what's happening. The idea that gamers care about "artistic intent" is hilarious if you've ever seen any gamer community.

Because "cheaper" isn't the only metric one uses to decide which type of power plant to build.

The #1 problem in all these discussions is that most people pick one single attribute, then say "Power plant type X is the best because it is the SELECT_ONE_OF( greenest | quickest to build | best ramp rate | cheapest to run | supplies the most power | safest)"

The underlying problem is that the laws protect the police so long as they *think* they are doing their duty. So it is almost impossible to bring a case against them.

No quality problem is ever solved by adding more signature lines to the paperwork. Code needs to be reviewed and tested no matter what the source. I just migrated some old software to a newer library, and used an automated tool to make some syntax changes to the code. Had it been an AI tool instead nothing would have gone differently: reviewed, tested, committed.

It is called the Department of War. Hilariously enough the rest of the Fortune article uses the correct name everywhere else, it's only the quoted sentence that misnames it.

Microsoft has tried this open approach many times, but they seem to always fail: Windows Phone, Zune / PlaysForSure, Media Center, Windows Tablet, Microsoft Surface, Windows Mixed Reality. Over the last 20 years people seem to prefer the walled garden over the open platform.

Who wants to speculate why?

There's a reason I phrased it "appearance of working" - you're assuming that enough people will be able to tell the difference between "working right" and "not working right." As long as it looks to be working properly for the majority of use cases, that's good enough. For most of these tasks, it isn't a simple binary between "doesn't work" and "does work," there's a whole spectrum.

In fact, I would argue this ultimately makes AI more dangerous, because it does a very good job of appearing to work while failing in ways a human doesn't.

Sure, probably, assuming it fails badly enough, which it might. But you're assuming "short term" won't be years, and that it fails in ways that make it clear profit was lost. It's pretty easy to assume that if a computer made a mistake, a human would have as well, especially if you're the one who put the computer in charge.

But that's the thing: AI doesn't have to work particularly well to displace hundreds of thousands of white collar jobs. It just has to create the appearance of working, while being cheaper.

It's already there in places where, even at minimum wage, it wouldn't be cost-effective to have a person perform the task, but an AI can do it cheaply enough. Even if it doesn't do it particularly well. That it can do it at all is enough.

Jobs are going to be given to AI, even if the AI does a worse job of it, simply because it's cheaper. The assumption will be that AIs will only ever get cheaper and more productive. The same assumption isn't being made of humans.