As with many metrics, it's easy to collect data, is reasonably accurate and most importantly, repeatable to identify trends; and of course totally useless except to see where you are burning money.

As hardware improves, local AI models will be much more common. Specific tasks, at TFA points out, don't need powerful LLMs. Local models that are much more task specific and can have defined rules can met the needs of many users; with an added benefit of keeping data local and private. Local AI isn't as fast or powerful, but will suffice for many tasks at a fraction of the cost. Of course, that means the days of free model d/ls may be numbered, since that would mean giving away the IP and reducing the use of paid cloud models; I suspect d/led models will at some point be history and replaced with subscriptions.

a typical case of American blind justice,

Or actually tried to answer the question.

Yea, our local high school has told teachers not to fail anyone. Maybe I should teach as a retirement job - at the start of the semester, tell teh kids homework is optional, two tests, pass everyone with at least a B and collect a check. Seems taht what the administration wants the teachers to do.

Running on WinARM could open up silicon Macs with a VM as a market as well.

I do the same unless I need it faster and the price delta is small enough.

Those products are a significant percentage of its business, they basically took the street vendors and gave them a global marketplace. You no longer need to go to Asia to get a high quality counterfeit football jersey or shoddy electrical appliance. It's even easier for the vendor, get cutoff and open under a new name. No new storefront to find, no goods to mov or hide from spot checks when the government decides to 'crack down' due to pressure from manufacturers. The quality varies greatly depending on product category, the bolts, screws, random hardware to use in 3D printed stuff, screen protectors, yard flags etc. were all fine and a fraction of Amazon's cost from a reshipper.

Amazon's actions will be interesting because it seems pretty clear that some of the sellers are just drop shipping from the same sources in China and marking it up to cover Amazon's costs and shipping.

If you needed insurance or healthcare were you able to obtain it?

Get your government hands off my Medicaid!

Fixed that for you.

I've always contended had he simply said "Yea, I got a blowjob in the Oval Office. You would have too if yo had the chance..." the whole thing would have blown over and never come up again. Nixon probably could have survived if he hadn't tried to cover up Watergate, served out his term and pardon the whole crew.

This isn’t just Sony. The original rights owners decided not go renew licenses, and Sony had no choice but to yank them or likely pay far more for continued access than they were worth. Any purchase should include a permanent non irrevocable right to access the content so customers companies like Sony can still legally provide access to them even if they lose the fights to sell or stream it. If the seller goes out of business or shutters servers, they have a right to a downloadable digital copy.

I agree, and such as mistake could easily end wrong, as the officers pointed out. I am not excusing the mistake made by the police, but there are a few compounding errors that made this happen. First - no check if it is a valid tag number on entry. You'd think Flock would at least do that, as well has have some pattern recognition to pull up an error when the entered number does not match the expected sequence to flag it for review. Secondly, no check to see how often taht tag number pops up and where so if it gets a hit the same day a thousand miles away lag it for review. CA only allows stolen for missing tags. I don't know what info Flock provides, but giving make/model/color with the tag could also help prevent errors since the cops would realize the stolen tag matched the car using it; although depending on how manufacturer tags are used it may not have helped in this particular situation. This is the classic cascading failures that resulted in the outcome.

A pro also isn't going to HD with his wife. Thefts by pros disappear quickly because, well, they are pros and want to avoid getting caught.

The problem, as shown in TFA, is the 10 are 2 small numbers stacked vertically between the 34 and DTM, so they get overlooked. Should the police looked closer, sure, but I can also see why the made the error because the 34 DTM is in much larger font size.

Yes, and should be legally liable for damages in cases like this. At. minimum, if there system catches 34 DTM in multiple areas at the same time, that's signs of a problem, and the art of pattern a computer should be good at detecting. The inability to report a tag as lost also means for this edge case it gets a stolen marker. You could tag it as lost and inform the tag owner not to reuse it if found, and if the correct number was entered if it shows up on a vehicle again mark it as stolen. That of course, would require work and changing a system.

From TFA:

Seems Apple has what they may need from their own systems. Given Apple's penchant for secrecy, they should have known Apple could see what tehy did and use it against them and OpenAI. I would also not be surprised if some employees who interviewed but didn't leave Apple told the approbate people was OpenAI was asking or in interviews.

What I don't get if you're a long term Apple employee at a relatively senior level, why risk everything you've made in stock alone by doing something that risks losing everything or a big chunk of it in a settlement? OpenAI could through them under the bus to save the company if needed, after all.

The difficulty would depend on what data the camera provides, analysis of traffic past the camera and available information about the locations of potential interest. A serious effort would be more than hacking cameras, it would include on the ground surveillance to collect information on route terrain, what identifying information cameras openly broadcast, correlating that with camera information and footage collected, etc. Government cameras such as traffic cameras might be a better target since they would likely be on areas of interest already and fewer to review.

Machine learning tools could make it somewhat easier to sort results into cameras of interest. Of course, it's a huge data set to go through if you want useful information. Depending on how many cameras you accessed you could just have people look at them and decide that way.

Is this some intelligence operation? Who knows. It's not cheap nor easy to get useful information, but something a state actor could certainly attempt to do if they want to dedicate the resources to it.