dpark

Fernandez deserved each of his yellow cards. Especially the second for that reckless and dangerous tackle.

Cucurella did not violate the rule. He touched his mouth for an instant. The rule is not “if you ever touch your mouth you get ejected”, which is why VAR didn’t do anything and neither did the ref on the field.

I generally have nothing against Argentina but they were wildly outplayed in the last game, they were disrespectful, and Messi’s appeal to the ref to try to get Cucurella ejected was shameful.

> Now, you seen more countries qualify, and you see countries who used to not qualify, dropping ASAP (in the deathmatch 16th matches, right after poule).

> So what did we get out of that? A few matches during the tournament, and they were quite frankly meaningless.

The answer for FIFA is of course money. More countries, more games, more money.

But the non-cynical answer is that all of those countries got to see their teams play 3 World Cup games. Some got to see more than that. I would bet that many of those countries’ citizens are very happy with the expanded tournament, and good for them. We also got Cape Verde this year, who many people around the world were excited for.

The NFL is so balanced not because it’s a “super” league but because every team has the same salary cap. The best team and the worst team all have to spend essentially the exact same amount of money on their players, ensuring that no team can buy their way to the top.

Even there it’s still not really true that the best team can lose to the worst team in any given night. The best team is generally favored a lot and justifiably so. In soccer teams can also be quite mismatched and a lucky goal by the worse team can be the only score and win the match.

I really hope AI is going to help turn the tide from this “everything runs a copy of chrome” architecture. It’s never been cheaper to build fast native UIs than right now, but everyone is still shipping web apps wrapped in chromium.

This is getting way into the weeds but the fact that costly signals are more trustworthy doesn’t mean all signals are costly.

An “in this house we believe” sign is signaling but if you live in a liberal neighborhood it’s extremely low cost by all measures.

They aren’t trained on compressed plaintext so I’m not sure of the relevance there. But regardless it’s my understanding that’s modern models are trained with orders of magnitude more storage than their parameters require. But it’s possible I’m incorrect. This is getting to the fringe of my knowledge of concrete LLM details.

These models are trained on way more than just books. GPT-3 was trained on about half a terabyte of filtered plaintext and the training corpuses have grown significantly by then by all accounts.

The books in question are not antique. The Twitter post makes this claim but so far as I can tell it’s not based in fact.

404 Media published a story about this as well and cites a bookseller who notes that all of the books there have sold have had ISBNs (and are thus from 1967 or later and generally would have active copyright).

”very large purchases were of books that had little in common, except for the fact that they all had ISBNs. This seller also sells rare books that do not have ISBNs, and none of those were part of the bulk purchases”

https://archive.ph/9MQrK

That would be a clear case of copyright infringement under current law. You can’t make a copy of a book and then give the original to someone else.

> When people call something virtue signalling what they usually mean is that it's cheap talk, but this is wrong. The whole point of signalling is that it's costly.

What? Signaling is not costly. It’s literally a display, a signal.

> That makes it even worse, then. This proves the original point.

I don’t know what the “original point” is here, but these AI companies are not providing “book excerpt services” and do not claim to. ChatGPT at least will refuse to provide detailed book excerpts (I hit a week or two ago myself).

> If we're building black and white straw man arguments, then sure, let's not archive anything.

It seems like you are the one creating the straw man. Do you have evidence that these companies are shredding actually rare books? The only cited concrete examples (in this thread anyway) are all rather boring. I seriously doubt they are shredding 200 year old books because why would they?

> they're using the info to regurgitate in some fashion and serve back.

Sure, in the same sense that they regurgitate any other text they consume. LLMs by definition do not have the full training dataset available, though. It’s far larger than the resulting model. So they can’t reliably reproduce full text without an external source (or if it’s in the training data repeatedly). ChatGPT actually refused to give me a bible quote the other day, presumably because I ran into some general “book regurgitation” safety net.

> The only expectation I have is for companies or anybody else to not destroy rare books. Is that such a tall order?

Honestly, yeah. The idea that people or corporations should hold onto books forever because of a cultural “ick” about throwing out books is a bit ridiculous. Most books end up in landfills.

They aren’t feeding Da Vinci manuscripts into this pipeline. They are feeding still-in-copyright books.

I don’t trust or expect AI companies to serve books at all. That’s not what they are scanning them for.

> being digitised so they can be found in electronic searches

You make it sound like they are running a second Project Gutenberg. They most definitely are not making these available for electronic searches. At least not searches the public can participate in.

That’s cool but also definitely not the same thing that GP claimed.

> * Everyone who isn't a FAANG or Frontier Lab seem to be absolutely starving for good talent.*

I think this is largely a pay gap problem. There’s a bit of prestige involved (why work for random corp’s dev dept when you could work at FAANG) but a huge part is that most those random corps are paying way less. The best talent generally goes where the money is.

> I don't like I'm pushed to migrate only because the software wants that, otherwise the hardware works with no problem.

This is one of the problems with the switch from shrink-wrap software to always-upgrading software. Customers feel like they bought a thing so it should work forever but what they are really asking is that new software be continually written for the hardware they bought years ago.

If you want to run your old hardware with your old software forever, you can. But of course in a connected world where a million malicious parties are trying to take over your system, this is probably a bad decision, and that’s where the tension arises.

> "Scientists bad because I don't understand statistics" is boring. Just because you decided that is a meaningless variable does not mean that it is.

It’s literally a comic about how spurious results arise.

I understand what you are trying to get at with your insistence on pedantically misreading the comic, but you’re missing the forest for the trees.

> Your null hypothesis is rejected, you publish and continue trying to explain why.

In this case the “why” is “we ran so many experiments that a false positive was inevitable”.

> Otherwise you stop and try something different.

“Only publish the positive results” is exactly the problem. This is terrible science because it magnifies false positives and hides the negatives (false or not). Do enough experiments and eventually you get a positive result for basically anything.

The annoyance of every AI announcing is a machine every time it’s invoked is obvious. I don’t see any basis offered for this actually doing anything beneficial.

> This is exactly my point. They are not repeating experiments. Each one is a different one, and independent of the other, because the chemicals in each color are different. They are not testing what you think that they are testing.

They ran the same experiment 20 times with a meaningless variable changed until they got a positive result with a confidence of 95%. (i.e. 1/20 chance of a random positive result.)

The point of the comic is that this is bad science. Or at least it is ver bad science reporting. This kind of nonsense is why there is a multibillion dollar market for supplements that do not hold up to further testing.

> This is not how science works in the real life.

True. There is a lot of bad science in real life.

The downside is it’s annoying as hell. The upside seems limited to a bureaucratic pat on the back for “doing something good”.

This sounds like the kind of well-intentioned thinking that results in products that are no safer practically but are far more unpleasant to interact with. This is like the pop up every time I start my car that reminds me I should not play with the screen while I’m driving.

> If they found a link, they found a link

You could have skipped the wall of text. They didn’t find a link. They found a random correlation.

The scientists here essentially engaged in P-hacking: https://en.wikipedia.org/wiki/Data_dredging

Munroe didn’t misunderstand the independence of the experiments at all. He understands that misleading spurious results can arise from repeated experiments.

> And it needs further research. Not less, not more.

No it doesn’t. At least not anymore than the other colors. A positive result is no more deserving of further investigation than a negative result. What you are describing is bad science.

It’s got electrolytes.

> I like motion, I wouldn't have set this myself!

You sure about that? “Turn on Reduce Motion” was a common bit of advice for people who were looking for ways to make Liquid Glass livable.

This seems deeply delusional at this point.

It’s obvious from context that his team owns some sort of WYSIWYG editor.

How was any of that incomprehensible? They gave a list of some actual things they shipped. The hand wavy dismissal where you pretend they didn’t say anything just exposes your bias. You don’t want to hear a success story so you don’t.

> They are kinda famous for systemic cheating at this point.

Is that really a Norway thing or a professional sports thing? Think of all the disqualifications in the Tour de France spanning many nations.

It’s not incompatible with open source. It’s incompatible with free software. If Apple or Google or Microsoft or the government needs to “bless” your build, then you have no freedom to actually use your build.