Latty

I've done AV1 encodes where I was convinced the conversion must have failed in some way for it to be that small, scrubbed through it and it all looks fine, watched through the entire thing to confirm that wasn't the case, just waiting for the shoe to drop, but then to find out, no, it's really just that good.

It makes a big difference to if you view LLMs as a tool that's going to do all the things the CEOs are claiming, or a tool that's in the middle of a bubble and is going to end up being a lot less than they claim (which can still be a useful tool, to be clear, they are claiming a lot).

You are right that there is nothing wrong with, e.g: Red Bull sponsoring an extreme athlete to do some crazy sport that was otherwise never going to get funding. It's true that otherwise they were never going to get funding for it, but it's advertising for Red Bull and something cool for people doing something interesting, win-win.

I think the difference is that Red Bull have a very strong income stream that funds it. Currently all these AI companies are burning investor money, and to me, that looks less like "savvy advertising" and more like "pumping the bubble".

They are problems humans couldn't solve given the level of interest and investment they had. You can believe the LLMs made it viable to solve them cheaply, that's the narrative they are selling, but I think it is not unreasonable to point out the companies are willing to burn way more money than the problems have been considered "worth" historically solving them to advertise their LLMs.

Finite doesn't mean small. We have no idea how much they are burning in the back to do this. If you are taking things people just don't care about enough to invest the necessary effort/money to solve, and spend way more than anyone else is willing to spend to solve them, the solution could be less AI and more access to lots of investor money you can burn on PR.

The real question is "how much would it cost me to solve a similarly hard problem?" which is impossible to answer without them giving the detail on how many of these projects they are running and how many tokens they are burning, something they don't tell us.

Or, less favourably: they are burning a ton of tokens on any random "hard" tasks they can think of to try and generate stories about AI doing amazing things to pump the hype, infinite monkeys theorem as a playbook. You only advertise the ones you succeed on, thus pushing the idea that's what they do routinely. Stochastic "genius".

Oned maybe.

I'm aware of the definition of censorship, but the context is you saying:

> One of the worst cultural developments of this century has been creative people becoming censorship-hungry prudes.

This is not the same as a dictionary definition of censorship generally.

Presenting people changing their views over time as they learn as some evidence of some new trend of "censorship-hungry prudes" is silly. It's hysteria and a false narrative trying to paint something normal and healthy as equivalent to something dangerous.

The much bigger media preservation issue is actually represented by Scrubs: the fact that all the streaming services now have the show with all the music cut out and replaced with generic royalty-free stuff because the licenses ran out.

Bill Lawrence removed episodes because he and the cast hated that for years after they'd see people at halloween dressing as the characters in blackface, and realised that people were taking the wrong message from the episodes they'd made which were meant to mock the idea of people thinking it was OK to do blackface.

That isn't self-censorship, it's accepting that your work didn't read the way you wanted it to and deciding it's not worth having it out there under your name any more.

Freedom of speech also includes the freedom not to speak, and people should be allowed to say "I don't like this work of mine any more, I'm not continuing to publish it". If you advocate for enforced speech, you are the censor, just using a different tool.

It's not a "skill issue", the product got worse.

If I wanted a chatbot, I'd go to a chatbot, I want a search. Making me frame my query as a question rather than just assuming it's a query by default is just bad.

Huh? It's extremely common for businesses to decide that breaking the law is a cost of doing business and just do it because they figure they'll end up net positive from it, or even just to cash out in the short term.

Uber made no attempt to cover up that they were operating without licenses, and just ate it and fought it betting they'd get established before the law could catch up, and they won that bet, paying some fines and stuff but ultimately taking the market.

These LLMs are literally trained by these companies pirating literally every bit of media humanity has ever made, they made very little attempt to cover it up.

Of course they'd be willing to break the law for some PR? With a thin layer of plausible deniability "oh no, we didn't mean for it to hack stuff!" they know they'll get a slap on the wrists at worst, all while generating hype to prop up the AI bubble further by presenting the models as hypercapable.

The mindset is probably: either a) the models become capable of what we are claiming and so the companies become so huge and valuable the cost is irrelevant and we'll be to big to punish meaningfully, or b) it's a bubble and might as well push it up as big as it can go while I can make money, then by the time consequences come around I'll be long gone and who cares.

I don't think they intentionally set it up to hack stuff with a prompt saying "hack this site".

I do think it's highly likely they knew this would happen with the lack of safeguards and number of instances of this stuff they were setting up, and that it's PR they want to make the models seem "powerful". Stochastic "unexpected" events they can advertise.

I suspect it was probably set up with the official internal goal of just trying a ton of arbitrary tasks that seem hard so that when any of them succeed they can publicise it and pretend the models do that routinely, but a "failure" where they hack stuff works just as well, if not better, for their goals.

Have you tried disabling the Page Visibility API? Web apps will sometimes use that to try and mess with background playing, IIRC.

Why would it be that is the only recovery method? The most convenient one is just more passkeys on other devices of course, but you can have any other recovery methods you want. Magic link emails are the most common, but there is nothing about passkeys stopping you from implementing anything you want.

Right, I haven't seen that particular flaw in baked-into-the-browser ones (probably because they already having secure domain matching code ready to use), but the distinction is that doing the domain check is a part of the spec and standard, so e.g: sites aren't going to break it by changing the log-in subdomain routinely, as has been common historically.

Password managers have had to be permissive enough to work with most websites, and there is no standard for it. There have been sites that blocked the autofilling of passwords and so on as well.

My point wasn't this one particular flaw in some password managers is the reason to use passkeys (the copy/paste point is the much bigger issue anyway), just that it's an example of how relatively brittle the password manager process is. Having it a core part of the spec gives stronger guarantees.