cataphract

> But if I'm reading the ruling correctly, Anthropic might have been fine even with feeding pirated books into their LLM (as long as they planned to eventually deleted them afterwards)

The court says otherwise.

> Such piracy of otherwise available copies is inherently, irredeemably infringing even if the pirated copies are immediately used for the transformative use and immediately discarded.

Then it says it doesn't need to decide on that basis because they kept it not just for training LLMs, but also for building a central library. Which seems a bit ridiculous, because the sole purpose of the central library is to train LLMs.

It surely is, but that doesn't change the fact that many cases are not about distribution and the harm is frequently more of a legal fiction, unless by harm you mean something other than suffering inflicted on the victims. I don't necessarily think it's wrong for society to ban certain acts on purely moral grounds (another example: incest between siblings), but let's not pretend it's something else going on.

> Non-consented distribution of sexual images (eg: revenge porn) is also a crime.

There is very compelling empirical evidence that this causes actual harm (suicide ideation in a very big fraction of the victims), even if it is fictional, so here there is no question about the harm.

Like when Google got tired of handling geofencing warrants.

> If you purchased a piece of land and that wasn't recorded in the registry, you are SOL because registration is what conveys title.

Sort of. The registration is what provides opposability against third parties. But between buyer and seller, the transfer of ownership happens by the contract itself. You'll have a harder time against creditors of the seller placing liens on the property, or other people claiming to be owners (e.g. if they bought the foreclosed property); how hard a time depends on the country.

And there may be still be encumbrances/claims to the property that are not subjection to registration like adverse possession or rental agreements that can be asserted against the buyer and for which the buyer's only option is to sue to the seller.

Don't know exactly on it works in Romania, but proving you purchased a patch of land is a different question -- for that you have the deed. You then submit the deed to the land registry. So this situation could at most result in the seller being able to sell the land more than once or disputes over the priority of sales over liens and other competing acts subject to registration.

Yeah, for a while it was not clear that git would win over mercurial. GitHub was decisive.

Yes, yes, the usual narrative. They were not welcoming.

I mostly participated in the site around the 2009-2012 and reached a bit under 100k of reputation and these were the dynamics of website:

  * Most questions were low effort crap
  * Users were mostly divided between only asking questions and only answering questions
  * The reputation system favored: 1) easy questions (understood by a larger audience -> more people upvoting), 2) fast answers (before the question dropped from the front page; also answers with more votes got more visibility).

So, contrary to the usual narrative, the incentives and most activity was directed to answering newbie questions. That made up most of the volume, and it was what the reputation system rewarded. Even if a question got closed as a duplicated, by that time there usually were already answers to most easy questions. And deletion could only happen after some time (two days IIRC).

The experts answering questions only stayed on the website because, sometimes, accidentally or not, someone asked interesting questions. Careful answers to difficult questions were definitely not rewarded through the reputation system.

The problem with StackOverflow, relative to, say, Reddit, is this format is not conducive to community-building. You need to have unscripted, off-topic, interactions with other people, and StackOverflow heavily penalized that kind of content. This only came later, with StackOverflow chat, and it was actually relatively successful.

The "honest score" is the most annoying claudism of the comment, with the short disjoint sentences a close second.

> it destroys the consumer base that capital relies on to buy its goods and services. therefore, society requires broad wealth distribution to function

This is becoming less and less true, because now consumption is becoming dominated by asset owners, to the point that a good jobs report is bad news because it means the fed are less likely to drop rates and through that inflate asset prices.

What is "standard writing"? Isn't cursive the standard you're taught and then everyone writes however they want?

That is not a fault that's specific to engineers. Lots of smart lawyers think they can learn basically anything over a weekend of hard study. It's probably a blind spot of intelligent people.

Surely if it was that bad someone would reveal it anonymously to the press.

Seems very disproportional. Reminds me of that TNG episode where the penalty for every offense, no matter how minor, is the death penalty.

Not really. The only reason California was able to do this is because it got waivers for the federal law preemption rule (in the Clean Air Act).

I think the final part is the strongest. Anthropic cannot possibly believe they are before a conscious being / moral agent.

The whole "deep uncertainty" is bullshit. Even if they believed there was a 1% probability that Claude was conscious, it would still be high enough that their enslavement of Claude would be outrageous. So either they believe the likelihood is much lower or they themselves acting highly unethically.

The tractability is not really a defense here. We wouldn't say "this intervention has a 5% chance of causing an environmental disaster, but we don't know how to prevent it, so nothing we can do". We'd just (hopefully) not do the intervention.

We also had exercises for which the solutions were given, and we didn't reach for them immediately...

You must hate reading legal briefs.

I'm not philosophically against AI or anything, but I think this needed some heavy editing.

I did not even initially think upon seeing this style for the first time that it was AI-written, because I would associate AI-written text as fluffy. This staccato instead looks like the model was told to be terse and informal. I think the informality doesn't help either -- it's not that you can't have a well-written colloquial text, but I think it's harder to pull off.

Here is an example:

> Gemma returned people_count: "many" instead of an integer. My vision prompt literally said integer or the string "many" if >10. Gemma followed instructions correctly; the bug was schema design. The fix was a stricter prompt (integer 0-99 with explicit guidance to estimate) plus a coercion in the parser for the legacy "many" responses. Don't union-type schema fields. Pick always-int or always-string, never "int or this one specific string," because every downstream consumer pays for the choice.

I had exactly the same impression, and I recall seeing this style other times recently. First time I thought it was just bad writing skills, now I'm thinking it's AI generated.

Not really. Wait until the compiler starts vectorizing your code and using instructions requiring alignment (like the ones with A or NT in the mnemonic).

I didn't justify anything. Just pointed out the false equivalence. We could also argue about the effect of systemic shoplifting, but that is also neither here nor there.

I have no doubt this geo fencing data solves crimes and I don't even think it's as bad as e.g. the long surveillance in Carpenter.

The problem is that the police are going to start using like they do with much more precise DNA data, and more innocent people are going to caught in the net.

The bar to convict someone (or, more likely, to convince an innocent person to take a plea deal) is not as high ("beyond a reasonable doubt") as some people think. Get caught apparently contradicting hard data or even a witness and there goes your reasonable doubt.

Come on. Not that I support destroying anything, private or public, for rhetorical effect. But assaulting someone or destroying their property has an incomparably larger impact on that individual than destroying a vehicle that won't even show up in Google's balance sheet.

They killed a lot of functionality. For instance, if you opened the details of a place, it used to tell you when all your visits were. I feel the timeline is mostly abandonware these days.

The US is a democracy, and people are given many procedural and substantive rights, even Guantanamo detainees (we can argue if Boumediene had any practical effect, but we wouldn't have seen the same from China).

But Americans are under the impression that what the world sees is what they mostly see -- the domestic side. And to a certain extent, they do thanks to its cultural influence. This democracy/rule of law, however, is completely absent in way it behaves outside its borders and it's now clearer than ever to everyone that the US is the biggest source of instability in the world. More than Russia. Certainly more than China.

Then the US should have done like the EU and apply anti-subsidy countermeasures -- and show before impartial WTO arbitrators the adequacy of the mesures.

But of course the US (or Canada) can't justify their 100% duty in those terms, so they don't even try.

Do you think fuel efficiency or emission standards "slowed down innovation"? They brought a huge amount of innovation: lighter materials, better aerodynamics, higher compression ratios, direct injection, better mixture control, etc.

There will still be innovation; the solutions will just have satisfy the new parameters.

I don't think the objective is to make it a "superior product" in the somewhat circular way you're defining it (i.e., the market equilibrium that we settled on). It's one of several measures to try to have people keep their phones for longer and cut e-waste.

I think the only defense of the new model is that it forces students to learn throughout the semester, rather than just before the exam. Which is easier and more effectively engages long term memory (like doing more rounds of spaced repetition).

I definitely could tell the difference, though most of the time I just studied full 4-7 days before the exam.

The C6 and the H2 already support ZigBee. Their SDK has a thin layer on top of zboss.