Let's train our AI on a bunch of fiction where AI goes crazy and kills everyone! What could possibly go wrong?

Whose bright idea was it to store nuclear waste right next to an ocean coastline?

As I said... not acceptable, but not necessarily nefarious.

I'm all for an arms-length agency that controls police data and equipment so there's no opportunity for people within an agency to cover each other's asses.

I think when you put on a badge and assume the authority it grants over others, you should be held to a much higher standard. This does not seem to be a common enough belief to actual have such standards implemented.

It's unacceptable, but not necessarily nefarious.

You have a case, you have a case number. The interface is stupid and you can't paste it in, you have to type it out every time. And every selected camera is another query with that same field waiting for the 15 character code.

You get bored, you type in crap knowing A) people rarely check if there's no complaint and B) you have a case.

I can't say that's what happens with every system, but that's why the video logs look like this on the ones I've touched. Which is also why we captured logon ID and didn't just rely on trusting the manual input.

You know what they fucked with far less often? CPIC queries. Abuse CPIC and you lose your job because if you don't lose your job, the RCMP pulls the police service's access. It should be the same locally, but it typically isn't.

Obviously, I agree.

My other question, is how long will they be able to get away with this sort of thing, before customers decide the risk of disclosing your data to these external AI firms, is greater than the cost of building the capacity to run your own models in-house.

I don't use LLM, generally. But I was messing around and got some older models running on my laptop last night. Nothing special, no idea what I would use them for (couldn't generate a paragraph of text before clearing losing all coherence, because these were small old models running on 5 year old hardware). Supposedly Mac Studios and Mac minis are selling like hotcakes because hobbyists want to run local models. How long before the Venn Diagram for "can be run locally" and "capable enough model" overlaps for most use cases, making the risk of data exfiltration by these firms not worth the risk?

Lot of words to say this kind of thing may accelerate the bubble bursting.

How many of the other miraculous math solutions attributed to LLMs recently were similarly snooped from academics who bowed to pressure to stay silent on the matter?

Nope. No matter where you look in the sky, on large enough scales it looks the same. There's no indication that things are significantly different in any particular direction.

There can't be a 'source' of a Big Bang that looks like that. It has to happen everywhere or the sky would show signs that it was all accelerating away from a single point.

What we can say is we don't really have much ability to figure out what happened early on. The conditions are too extreme and we don't have any way to model what might have been going on. That's not faith either, that's, "We don't know and might never be able to figure it out".

But note how when it comes to copyright infringement claims suddenly the argument changes to "but it learns like a real person so that's protected by fair use laws!" Somehow, somewhere, someone in charge of actually enforcing the law has to realize we can't let them have it both ways.

No European search engines to choose from?

There are, tried them, but my conclusion is that country of origin doesn't seem to matter much. I rarely search these days but if I do I use my hermes bot via telegram, that uses a few different search APIs underneath. That together with some reasonable instructions to the ai works fairly well.

They did at some point, years ago, quietly remove that clause from their company charter. There was no press release about it when it happened, but it was reported here on Slashdot at the time. IIRC, it was right around the same time they got caught illegally photographing private property and mapping wifi network SSIDs for their "Street View" data.

Maybe that's just wishful thinking though.

Maybe Brave respects my choices more, but I still install Ghostery and UBlock Origin.

>If this is true then wouldn't the Biden admin have wanted the BLS to hide or change the CPI reports during those first couple years in the supply chain crunch when those numbers were just destroying their approval ratings?

The difference is between 'spin' and 'outright lying'. Normal governments will show the stats you can examine until the cows come home without being able to say they're lying. They adjust the definition of who is employed or seeking employment, things like that. They can still look bad, because they're still reporting aspects of the real facts.

I think the measure I was talking about is referred to as the "Labor Force Participation Rate", but I'm not sure.

There's a stat - I forget the name for it - but it's basically "The percentage of hours worked out of all those who would normally be expected to be working".

If you're of working age, not prevented by disability, etc., you would be counted. It can't be fudged by not counting people who have just given up, it can't be fudged by counting part time employment. And you can combine that with the wage distribution curve, and then the wealth distribution curve to get a pretty good picture of the overall economic health of your population.

I've never heard of a government that wants that. Generally, economic reports are spun to make the government look more effective than it has been and to protect the wealthy against accusations of hoarding that might result in taxation. A really bad government might just give up on spin and, say, fire people who won't just tell whatever lie they want.

Not sure if this is low-quality astroturf, bad trolling or a failed attempt at sarcasm but that statement is objectively false.