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?

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.

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?

That is not true. There's no magical alien group who nominates candidates. We the People nominate them, and this is what We keep choosing

Apple's mistake was building *privacy* into their AI model. Nobody else did that, and it crippled Apple's solution. Apple pushed 5GB of AI model data to everyone's phone, and everyone complained about the space usage. Next up their AI is slow because it is using resources on the phone instead of big data centers. Apple did what everyone asked for, but users were ultimately unwilling to accept the compromise.

Personally, I liked the Apple solution better. On Android, if I lose internet for 2 seconds, and say to my phone "Call Bob Smith" it sits there for several minutes then times out with "try again later". BUT IT GETS WORSE: The local hardware transcribed the text perfectly. So there was no need for a server to be involved at all!

Apple's old approach was the right one, so it is really sad that they botched it.

EFF resources on age verification
I thought I understood the issues and options and trade-offs, but after reading this I realize there is much more to it than I considered. This is a really good resource, and some of the articles under it are bite-sized so you can send to non-techies and make them go "oh yeaah.... I hadn't thought about it that way.... maybe this isn't such a good idea....?!"

I am not the OP, but perhaps each vendor should not have to architect this function, since it is such a high security risk. Instead, use infrastructure like apt, yum, chocolatey, Windows Update, Steam, etc.

You are right. And also, the advertiser/data broker won't want the data for "only people who don't pay." They are going to be very interested in those people.

Many people avoid Meta products because of the predatory pricing model. In that model, "payment" means giving up our privacy and control of our data to advertisers and data brokers. Instead, offer a model where we pay directly, and give us control of our own data again.

Any other deal is pointless. Any "premium" service is for their benefit, not ours.