Right, I’ve never once had to fill a cars tank with gas, or change the oil, or pay for any other maintenance on one.

Hardware is Chromebook priced. OS,is (AFAIK) full macOS, AKA a posix compliant Unix machine with a pretty nice GUI. Nice enough that several Linux WMs try to duplicate it.

I guess google included the Buffy episode where a demon “AI” gets its followers to make it a body.

The government may not be able to bail these companies out. The scale is even bigger than the housing crisis of 2008, and trust in the current administration is basically zero. I think the most we can hope for is the LLM companies (think OpenAI and Anthropic), and the companies whose services are effectively wrappers for LLMs, and probably Oracle (with its negative cash flow and astronomical debt) all go away. Amazon, Microsoft, and Google probably survive, with some high profile bloodletting, senior executives being purged by their boards. Apple has been the least bullish on AI, so they’re probably more or less safe and the biggest change will be new OS versions that don’t refer to Apple Intelligence. Facebook is structured in such a way that Zucc can’t be removed by the board, so who knows how that plays out.

Palantir and their ilk will likely get whatever they need to survive unless the midterms bring in a shockingly progressive group that cares about people’s privacy and removes funding for mass surveillance.

Notably there have been almost zero data breaches of large banks, because their requirements for security are significantly higher than most other companies. My original comment was not about banks, they obviously need to retain a lot of customer data, and most of that is not exposed to the internet at all. I was talking about things like a pizza shop or an online retailer. There’s no need for Burger King or a webcomic artist I’m buying a print from to have a login or my email address for longer than it takes me to get my items.

Tax records don’t have to include the customer if it’s retail. If that was a requirement cash businesses would have massive problems, and the rule of keeping those records for seven years significantly predates our current model of credit for everything.

Beyond that, if I go to a restaurant they don’t have my name and address or any other information. Businesses need to keep records like “we bought x from y for $z,” and “we sold x to a for $b.”

And even further, the government could clarify that (if in some countries customer data was part of tax data) that the law was now to protect customer privacy and data.

The EU GDPR doesn’t go nearly far enough.

If I order online, my data only needs to be retained until I get my item. A electronic receipt can be sent via email.

Social networks should have human moderation, and not insist on retaining real-world data about users.

These things could be accomplished through regulation, and if enough countries (or US states) put those regulations in place it will eventually be more cost-effective for companies to implement the changes globally.

The Pitt is covering this, and in an early episode from this season they had one of the doctors point out that the LLM transcription incorrectly labeled a medication.

Medicine has a very low tolerance for errors. If I ask ChatGPT what episode of Downton Abbey shows lord whatshisface vomiting blood and it tells me that episode was the Red Wedding, worst case scenario is I look dumb. If Claude tells a doctor “this patient doesn’t have any existing medications that are contraindicated for propofol,” and it’s wrong, that patient may die on the table.