For the purposes of this bill - https://fastdemocracy.com/bill... - as far as I can see it seems neither federal nor California "exempt" status is relevant.

The question is malformed. The set of "microplastics" includes both things that are a problem to have in the body, and things that are not.
The question is akin to "What solutions do you suggest to prevent fluids from ending up in the human body".
Where specific plastics are identified to have specific harms, and the mechanisms understood, then specific measures can be taken to address those cases.

This looks almost like a more serious implementation of INTERCAL's COME FROM or the CLC-INTERCAL extension of NEXT FROM/RESUME. A reverse goto/return elsewhere that triggers on a line which has no obvious indication that it triggers anything.

Obviously the AI training process identified some pattern with predictive power in the training data that humans had missed. So now since we know a pattern exists, and is a simpler approximation than simulating and predicting the full physics in detail, they can go and look at what the AI did and the associations it made to create a clean heuristic for predicting these things in future. That could well provide a powerful component in modeling tools.
If they just stop here, at letting AI do it for us, no one ever actually learns from the experience.

Adblock is very different. Adblock is someone choosing not to download some parts of something, rather than downloading something from a system the owner of the system has said their not allowed to download from.
Robots.txt is the defacto standard, the industry norm, for limiting or denying automated systems access to servers.
And yes, if you go on a website and it has a notice you're only allowed to visit the website if you are a member of group X (eg, employees) then that can be legally binding too (albeit easy to bypass in practice).

Trespassing is just an analogy here. Unauthorised access is the actual issue. Robots.txt is a industry norm for advising whether an automated system is permitted to have access to the system. Generally, you don't have to have a substantive enforcement mechanism for such a prohibition to be legally binding - in the absence of permission you can't access that system.

A "No girls allowed" sign on a treehouse might, if the person putting the sign up had ownership rights, be enforceable against those who violate the sign as being trespassers though.

Robots.txt does serve as a kind of "no trespassing" sign, and it'd be interesting to see how it holds up in court in terms of serving as commonly recognized notice of limitations on permission to access a service, for automated systems. Quite often just putting someone on notice that they're not allowed to access something is sufficient to turn it into a crime, and robots.txt is how the internet community puts automated agents on notice.

The problem here isn't AI misalignments, it's people blindly trusting and relying on AI without thinking about it.
Don't do that.

Long Service leave (which is more commonly granted after 10 years service, and is usually about 8 weeks at that point) does seem somewhat specific to Australia - goes back to a tradition of people going back to visit the UK by ship, after working in the colonies a while.
But otherwise yeah.

Many of those permissions (the system privileges) are represented by flags on your authentication token, and a process can voluntarily turn off privileges it has while remaining in the same user process.
That's basically what UAC does for users with admin privileges - early on in the login process, the session drops privileges considered sensitive, and the UAC interstitial is what allows a process to add them back on again without having to do a full login from scratch.
Maybe sudo will just be a command line UAC flow for admin users in the end? An "elevate" command, basically? It's about all I can think of that makes sense.
Though how that will work with the security context for window interactions I'm not sure. Normally unprivileged processes are limited in how they can interact with privileged windows, not sure if Microsoft will be doing magic to elevate the attached conhost instance for the terminal along with the underlying shell?

More thinking something like Disney might find it cheaper to whack an author than lose a nine or ten figure lawsuit over rights.
A flat 25 years makes sense though sure. Moral rights absolutely should lapse at death, where they even exist (they're not universally accepted to begin with)
I agree with the absurdity over the Tolkien estate absolutely.

I agree with you generally, except that this would incentivise, for lack of a better term, "copyright assassination" - the deliberate termination of someone in order to release lucrative copyright.
A reasonable amount of time (a decade or less) after an author's passing should mitigate this for the most part, or tie copyright to expected lifespan (eg, "100 years after the author's birth")

This wouldn't apply to corporate-owned copyrights, of course. A hard time limit should be fine there because the interest in the copyright would always pass to someone

Those codes are presumably for the online test portals etc? If so, the problem is universities outsourcing their testing and scoring to the textbook providers.

You can specify "any edition" or a range of acceptable editions. Universities and schools alike once had thriving secondhand textbook markets, there's no reason (save for the aforementioned outsourcing) that there couldn't be again.

I believe it's the original article trying to be poetic while referring to the problems of an AI system intended to police human speech being used to police AI generated speech (now that LLM generated speech undergoing a rapid growth)

Sort of. If you control enough variables it can be. It's as deterministic as a PRNG would be. To get truly nondeterministic behaviour you have to bring random input in from external sources.