Assuming it's the US we're talking about, the federal minimum wage is $7.25, which means that if every worker involved is paid at the minimum wage, you incur a cent of labour costs every 4.97 person-seconds. AFAICT, most Amazon workers are paid substantially higher than the federal minimum wage. And that's just labour costs.

While Amazon is efficient, "fractions of a cent" is probably the wrong order of magnitude for even the most efficient order.

That's not even mentioning their additional overheads, like the cost of fuel for their idling van as they drop off your potato.

In Australia, if a company fails to pay its Goods and Services Tax, or its withholding payments for its employees' income tax, or Superannuation Guarantee (retirement fund contributions), each and every director is personally liable - unless they can prove (and the onus of proof is on them, not the Tax Office) that they had a reasonable excuse for why they weren't active in managing the company, or took all reasonable steps to get the company to either pay the debt or go into bankruptcy the lawful way.

Importantly, there is no "wilful" requirement and it applies to all directors, not just those who actively participated in misconduct. If you weren't involved, you have to prove that you actively tried to stop it, or that you weren't managing the company for a specific reason such as being sick. You were the director who mostly turned up to board meetings to help them meet quorum, you trusted the other directors on the board had things under control and you were completely unaware of the debts? Too bad, liable. You hired external advisers, delegated to them, and they didn't do it? Too bad! You decided to wash your hands of the whole thing, and resigned from the board, but didn't actively try to rectify the situation first? Yep, they're still coming for you.

(I believe the criminal convictions with prison time only really kick in for those who actively participated in tax fraud or who refuse to pay their director penalties.)

My initial reactions:

- I hope they succeed and eventually deliver a solid version of this product - verifiable photography is going to become important, and it's good to see startups working on this - While I'm sure some artists will like the idea of verifiable photography, the applications that matter to me are any kind of photography that has the potential to end up in a news article or in court - Selling what is essentially a prototype is fine, it's extremely obvious that's what it is, they explicitly say it! Who cares if it's not very good as a camera? - The almost complete lack of information on their site about their security model or how their ZKPs work is not particularly encouraging - It follows that my faith that either the cryptography or the hardware anti-tamper measures in this beta device would stand up to even some decent amateurs, given a couple of weeks to have a crack at it, is not high. I'm almost tempted to buy one just to see how far I, a random kernel engineer who gets modestly decent scores at my local hacker con CTF, could get. But I may well be completely underestimating them! Hard to tell with the fairly scarce information - Why did they pick a name that's similar to a) AMD's GPU stack, and b) the law enforcement/natsec computer vision business, ROC (https://roc.ai)?

I'd be shocked if the major sensor vendors don't already have engineers working on exactly that, though.

Sony has this - https://authenticity.sony.net/camera/en-us/index.html

It's bizarre to me, an Australian, how the pardon power is used in the US. Our federal, state and territory executive governments all have a pardon power, inherited from English law, that is, formally, unlimited (like the US federally and indeed it's less restrictive than many US states for state crimes).

It is a power used very sparingly, even though legally it is unlimited - the state of New South Wales is, as far as I know, the only one which publishes details about uses of the pardon power; in an average year there are 0 successful pardon/commutation applicants, and it's an exceptionally merciful year if they grant 2 or more. Other states and the federal government may or may not be a bit more generous, but we're talking very small numbers. Most pardons are for reasons of unsafe convictions where for whatever reason no remaining avenues of appeal are available (rare, these days, because each state has introduced laws to enable post-conviction reviews).

Historically, particularly in the 19th century convict era, the pardon power was much more important, and was indeed abused for political reasons on a number of occasions, but it seems that for the most part it quietly exists in the background and only gets significant public attention once every blue moon for a high-profile murder case or similar.

What explains the difference? Is it the requirement for sign-off by the King's viceroys that prevents abuse? Collective Cabinet governance that is accountable to Parliament? Maybe our political culture means politicians' friends tend to end up in prison less often and thus there's less opportunity for the abuse of pardons specifically? It's not particularly clear to me - if anyone's got some good comparative studies send me links!

Yeah, you're wrong - a court simply isn't going to consider a lawnmower's translation of throttle input to motor power as "inference". The principles of statutory interpretation require courts to consider the context and purpose of the legislation, and everyone knows this is about GPT-5, not lawnmowers.

In any case, that definition is only used to further define "foundation model": "an artificial intelligence model that is all of the following: (1) Trained on a broad data set. (2) Designed for generality of output. (3) Adaptable to a wide range of distinctive tasks." This legislation is very clearly not supposed to cover your average ML classifier.

A relative of mine had to go back to their paper-only specialist a couple of months ago to get a prescription reissued because the specialist had omitted a mandatory detail from the (handwritten) prescription form and the pharmacist couldn't fill it.

Meanwhile, I had a similar prescription, from a different specialist, who issues his prescriptions as either e-scripts or computer-generated paper scripts depending on patient preference. I suspect his practice management software would stop him from making this class of mistake entirely.

I get why a doctor might prefer to avoid the computer, but I think my relative would have preferred their doctor not screwing up on something basic and wasting a significant amount of their time over better vibes in a consult.

This is true, but the general advice one gives through, for example, publishing a style guide, should focus on producing reliable outcomes over great outcomes. I'd happily lose most of the great documentation I've read in the course of my work (not that much) for never having to read inaccurate or woefully incomplete documentation ever again.

Can't speak for the US National Archives, but I do occasionally head to the National Archives of Australia to look at things that pique my interest, and I can confirm that yes, you can just ask for entire folders full of documents and in most cases, they'll take the originals and just plop them in a plastic tub for you to pick up and have a read through. It is a slightly strange feeling opening up a file and flicking through correspondence that was once personally handled by a Prime Minister from 100+ years ago. (They will use scanned copies for anything that either can't be handled for conservation reasons or has been redacted for security reasons.)

Give them the credit after the sponsor actually finalises the updated contract.