1. require functioning parent controls (here its also worth considering the abuse of parent controls). A problem here is the absence of a proper competing market of (phone) operating systems...
2. use parent controls to "anchor"/set the age/age category. I.e. NOT age verification, just age indication.
3. propagate them similar but not the same as with the Californian law, where possible do the decision before starting a program/fully loading the website etc. (1)
4. allow exception to be set (incl. per "origin" i.e. app, but also sub app e.g. browser:<domain>), it's a parenting tool not a state enforced
5. make it explicitly a non goal for this to be "hard to hack" or anything like that, it's a parenting tool not a banking tool. Proper trust management in a parent child relationship still matters and replacing it with "technology" is unlikely to end up well.
6. where possible leave the decision to the parent controls
7. Age categories are geographically/standard/age scoped, for most apps/site they only have one age gate and can just list them, potentially with content group hints, e.g. `us:pg:13,horror;de:fsk:12,horror` if the the user is in idk. uk the parent controls can make the decision, which might involve parent settings. E.g. a German parent probably wants to treat `us:pg:13,violence` as 14+ and very conservative people in the US want to treat `de:fsk:12,non-erotic-nudity` as 16+. For apps which serve content on a feed it's more shitty as they really want to be given the age gate instead of providing the contents age on access. This doesn't mean that they can't check the age gate for every peace of content "when serving it" (pitching back control to the parents controls) but still need a general age category, which will leak parent controls country, most times that will happen anyway by IP country of origin. So should work?
8. IP country of origin != age gate law which should apply. While legally not fully wrong to treat the same parents would be very surprised if their parent control allow/forbid things when they are on a holiday trip or because their child connected to a VPN tunneled hot spot... This loops back to 6. to give the decisions to the parent controls instead of the app/site/service.
9. criminal liability for intentional miss-classification of age gates (the "intentional" part matter a lot here).
If you are found personally responsible for tax evasion >1e6€ then the minimal penalty is prison sentence without parole option. This is true for many EU countries including Italy. Idk. about the max. prison length in Italy for this but e.g. where I live in the EU you are likely looking at ~15 years for 1e9€ tax evasion.
The reason executives commonly avoid such penalties is because they avoid being found personally liable by claiming they didn't known, did misunderstood the situation, where deceived by others etc.
Through it should be noted that this case is a bit unusual and complicated.
The tax dispute itself isn't as simple as Amazone directly having avoided paying their own taxes. And the case of missing taxes has already been settled. This new current investigations are criminal investigation (i.e. the failure of paying taxes is assumed to have been intentional instead of a booking error) and seem to be more targeting executives for having committed crimes (instead of targeting Amazone the company).
Or in other words, Italian prosecutors are feed up US companies not caring for EU law and no one being hold liable.
---
(1): Without option to have it replaced with long time parole.
> The reason executives commonly avoid such penalties is because they avoid being found personally liable by claiming they didn't known, did misunderstood the situation, where deceived by others etc.
In the US, Section 302 of the Sarbanes-Oxley Act exists literally to avoid this: the CEO and CFO have to personally sign filings so they cannot "not know" about financial disclosure content and have plausible deniability. While it only applies to public companies, the model seems reasonable to solve this case too.
"not know" was oversimplified from me, technically such things exist in many countries, most likely also Italy.
But this comes back to how thinks always tend to be more complicate then "clean cut direct tax avoidance".
E.g. here the case isn't as simple as Amazone directly avoiding tax. But them instead to quote:
> Amazon's algorithm and operating models enabled the sale in Italy of goods from tens of thousands of non‑EU sellers - mostly Chinese - without disclosing their identity, helping them avoid paying value‑added tax (VAT)
I.e. Amazon is seen responsible for acting with sever levels of negligence with suspicion of intentionally enabling/supporting tax evasion to profit from it. (And honestly given the level of fraudulent-looking things Amazone allows vendors since a very long time, there is likely some truth to the "intentional supporting malicious actions" part.).
And in turn executives don't claim exactly "they didn't know". They claim they did know and started actions to fix the issue but it the actions failed so you tried other actions and if asked why no bigger actions are tried they claim it didn't look like it was "that" big of a problem through they now realize that was wrong.
The end effect is still the same they doge responsibility as-if they could just claim they didn't know, just with extra steps. Instead of "didn't know" I probably should have used "they claimed incompetence" (and external factors outside of their control) or similar.
> If you are found personally responsible for tax evasion >1e6€ then the minimal penalty is prison sentence without parole option. This is true for many EU countries including Italy. Idk. about the max. prison length in Italy for this but e.g. where I live in the EU you are likely looking at ~15 years for 1e9€ tax evasion.
There's o EU wide tax law, so that statement is misleading at best and there's many places where that's false. You're spreading misinformation.
> The reason executives commonly avoid such penalties is because they avoid being found personally liable by claiming they didn't known, did misunderstood the situation, where deceived by others etc.
Misinformation again. This blanket statement in just false. Again, because there is no EU-wide law. It's up to the countries and they handle it very differently.
AI can do a grate job for grammar, spell and formulation checking/fixing without changing any content. I.e. just adding as a fancy version of extended spell checking.
While I do currently not use it like that there shouldn't be any reason to ban it.
And tbh. given some recent comments I have been really wondering if I should use it, because either there are quite a bunch of people with lacking reading comprehension or quite a bunch of people with prejudice against people struggling with English spelling and grammar.
Either way using AI as extended spell checker does would help with getting the message through to both groups as
- it helps with spelling, grammar in ways where traditional spell checker fail hard
- it tends to recommend very easy to read sentence structure and information density
except you can nudge LLMs to use different stiles more similar to your writing
they aren't good at it but viable
and more important this is about LLMs fixing grammar, spelling and pointing out bad formulations with change recommendations. This is not about giving them pullet points and telling them to write text for you.
but spelling and grammar still isn't a good indicator for expertise, intelligence or anything like that even in an academic context
Mainly:
1. Dyslexia doesn't make you dump, just likely to misspell and a less likely to notice your misspelling.
2. When speaking about neurodivergence people mainly think about Autism or ADHD but sometimes just mean that your brain thinks in very different patters, this can make grammar hard. Especially if it's not your native language.
3. Sometimes people had shitty situations earlier in their live, leading to incorrectly learning parts of languages. This is hard to fix. But isn't really representative in any way for their expertise in any topic which isn't the given languages grammar.
4. English grammar and pronunciation to spelling mapping aren't exactly well designed. People not wanting to bother with it is not really related to intelligence, or excellence in other topics.
5. Some kinds of expertise are unrelated to general intelligence, expertise, education. So even if spelling and grammar where related to intelligence, it wouldn't be meaningful to judge expertise.
I think the grammar/spelling is just one (perhaps low-signal) sign. But a lot of these people really are not that intelligent. And not just the GlobalElite™. Think of the guy who owns the local car dealership or owns 20 laundromats in the surrounding 3 counties. These guys are not geniuses, either. They just happen to own things that make them rich.
I worked with a tech founder at one point in my life, and I once happened to get a glance at his undergrad college transcripts which were, for reasons unknown, just sitting out on his desk. It was all Ds and Cs. He barely graduated! Yet his networth was more than the combined net worth of all of his employees.
Your GPA isn't necessarily a measure of your intelligence. I graduated with a 2.01 GPA from college, because I spent most of my time learning about technology and things that interested me, and doing the bare minimum to pass my classes.
But my diploma still says "UC Berkeley" on it, just like the guys with the 3.9 GPA. And when I hang out with PhD friends' PhD friends, they just assume I'm a PhD too.
So what I'm saying is that sometimes smart people don't put a lot of effort into school.
I can't tell my kid with a straight face, "Work hard, study, get good grades in school, and focus on a good career" when I know it's fucking bullshit. And what I should be saying is "Sorry that I'm not rich and well connected--since that would have been the outsized predictor of your life success."
It is just also not an argument against their intellect either.
Just look at what they nonsense they say all the time, and how they arguments and reasoning is commonly full of hols and messy entangled problems often stump them.
Similar for many cases where it's publicly visible its very clear that their success story commonly highly relies on stuff like knowing the right people and luck of having the right think at the right time with the right supporter.
But it would be a mistake to assume that non of them are very intelligent (or that non are quite dump), that would run at risk of underestimating how dangerous they can be. Both in "clever dangerous" and "idiotic dangerous" ways.
The reason I care about this is because if you ignore the US for a moment in a lot of places the same kind of "all connections often little substance" elite exist, just with having bothered to learn to use "extra" eloquent language to the as arrogant if not even more arrogant look down on people.
Which brings us to another reason why it isn't a good indicator: In the same way that "elites" use eloquent language to differentiate them self form common people do the "elites" around the trump camp differentiate them-self by explicitly not using it.
like if EU countries themself decide to undermine their own rights thats their thing
but if external powers try to undermine EU constitutions isn't that a attack on those countries and shouldn't that be a crime, if not punished else wise at least punished by exclusion from lobbying and if not respected by exclusion of the EU market as a whole?
sure freedom of speech, including lobbying, is VERY important. But there is a big difference between external powers trying to take away rights of EU citizens and internal powers doing so and that is also quite different to what lobbying is made for (trying to avoid making laws which are "accidentally hurtful"(1)/have unintended side effects or are simple not viable to practically comply to(2)).
(1): Like actually accidentally/unintended harm, not a reasonable law some company can't comply with because their whole foundation is based on idk. customer abuse.
(2): Where it's not the point of the law. If a company makes money by causing mass harm through a legal loophole and that hole is closed it should never matter if they can comply with it.
---
Can we also please start a corruption investigation into politicians involved, politicians which have repeatably tried to undermine their own countries constitution (weather on country or EU level) shouldn't be allowed to just go one like nothing happened and try again a year later.
which never should have been a thing,
because it was obviously wrong
yes AIs is more likely to use em-dash,
but that is just one, by itself very insufficient, indicator.
it's like hip size. In average over the populations
they are wider for woman. But the effect is too small
to classify the gender of a hip bone by it's size.
(Like for a specific age range and ethnicity, the difference
in median is like 1" or so, while there is a >10" difference
between 5%-percentile and 95%-percentile. Varying by gender
in difference and exact distribution.) Well I guess em-dash
are more an indication for AI then hip size for gender... lol
never understood why -- => em-dash auto completion is only a think in some subset of application instead of being a standard behavior for (display) text inputs
maybe I should have called it "natural language text input"
text inputs for (non NLP) machines are a special case
and for some rare niche edge cases it's not like you can't "undo auto change" (if proper implemented)
still the default for all WYSIWYG editors and text field in browsers should be to do it (with the option to switch it off)
PS: Also in case it isn't clear it's `<(?:\w)>--<(?:\w)>` to `\1<em-dash>\2` auto completion. I.e. `---` for markdown isn't affected, nor is `i--` or `--i` or similar. It kind makes most "programming language" edge cases non problems.
Because the EM Dash is not universally used. In Germany we use the EN Dash. That's also why the proposal is dumb. It has the typical US-centric view you would expect from a typical American.
which is the actual relevant part: they didn't do that dance AFIK
AI is a tool, they set it up to make a non-verbatim copy of a program.
Then they feed it the original software (AFIK).
Which makes it a side by side copy, as in the original source was used as reference to create the new program. Which tend to be seen as derived work even if very different.
IMHO They would have to:
1. create a specification of the software _without looking at the source code_, i.e. by behavior observation (and an interface description). I.e. you give the AI access to running the program, but not to looking into the insides of it. I really don't think they did it as even with AI it's a huge pain as you normally can't just brute force all combinations of inputs and instead need to have a scientific model=>test=>refine loop (which AI can do, but can take long and get stuck, so you want it human assisted, and the human can't have inside knowledge about the program).
2. then generate a new program from specification, And only from it. No git history, no original source code access, no program access, no shared AI state or anything like that.
Also for the extra mile of legal risk avoidance do both human assisted and use unrelated 3rd parties without inside knowledge for both steps.
While this does majorly cut cost of a clean room approach, it still isn't cost free. And still is a legal mine field if done by a single person, especially if they have enough familiarity to potentially remember specific peaces of code verbatim.
Well sure they didn't do the dance, but you don't have to do the dance. The reason to do it is that it's a good defense in a lawsuit. Like you say, all of this is a legal minefield.
So my understanding was that the original code was specifically not fed into Claude. But was almost certainly part of its training data, which complicates things, but if that's fair use then it's not relevant? If training's not fair use and taints the output, then new-chardet is a derivative of a lot of things, not just old-chardet...
This is all new legal ground. I'm not sure if anyone will go to court over chardet, though, but something that's an actual money-maker or an FSF flagship project like readline, on the other hand, well that's a lot more likely.
Strong agree on it all being a legal minefield / new grass.
> But was almost certainly part of its training data, which complicates things
On this point specifically, my read of the Anthropic lawsuit was one of the precedents was that if it trains on something but does not regurgitate it, its fair use? Might help the argument that it was clean-room but ¯\_(ツ)_/¯
My understanding is they did do the dance. From the article: "He fed only the API and the test suite to Claude and asked it to reimplement the library from scratch."
One could still make the argument that using the test suite was a critical contributing factor, but it is not a part of the resulting library. So in my uninformed opinion, it seems to me like the clean room argument does apply.
Is your (1) description of clean room implementation, and (2) description of what was done, actually correct?
(1): my understanding was that a party _with access to copyrighted material_ made the functional spec, which was communicated to a party without access [1]. Under my understanding, theres no requirement for the authors of the functional spec to be 'clean'.
(2) Afaict, they limited the AI to access of just the functional spec and audited that it did not see the original source.
Edit: Not sure if sharing the 'test suite' matters, probably something for the courts in the unlikely event this ever gets there.
[1] Following the definition of clean room re implementation as it relates to US precedent, ie that described in the wikipedia page.
Which is IMHO my take on what should be done.
1. require functioning parent controls (here its also worth considering the abuse of parent controls). A problem here is the absence of a proper competing market of (phone) operating systems...
2. use parent controls to "anchor"/set the age/age category. I.e. NOT age verification, just age indication.
3. propagate them similar but not the same as with the Californian law, where possible do the decision before starting a program/fully loading the website etc. (1)
4. allow exception to be set (incl. per "origin" i.e. app, but also sub app e.g. browser:<domain>), it's a parenting tool not a state enforced
5. make it explicitly a non goal for this to be "hard to hack" or anything like that, it's a parenting tool not a banking tool. Proper trust management in a parent child relationship still matters and replacing it with "technology" is unlikely to end up well.
6. where possible leave the decision to the parent controls
7. Age categories are geographically/standard/age scoped, for most apps/site they only have one age gate and can just list them, potentially with content group hints, e.g. `us:pg:13,horror;de:fsk:12,horror` if the the user is in idk. uk the parent controls can make the decision, which might involve parent settings. E.g. a German parent probably wants to treat `us:pg:13,violence` as 14+ and very conservative people in the US want to treat `de:fsk:12,non-erotic-nudity` as 16+. For apps which serve content on a feed it's more shitty as they really want to be given the age gate instead of providing the contents age on access. This doesn't mean that they can't check the age gate for every peace of content "when serving it" (pitching back control to the parents controls) but still need a general age category, which will leak parent controls country, most times that will happen anyway by IP country of origin. So should work?
8. IP country of origin != age gate law which should apply. While legally not fully wrong to treat the same parents would be very surprised if their parent control allow/forbid things when they are on a holiday trip or because their child connected to a VPN tunneled hot spot... This loops back to 6. to give the decisions to the parent controls instead of the app/site/service.
9. criminal liability for intentional miss-classification of age gates (the "intentional" part matter a lot here).
reply