ticulatedspline

CSAM is highly illegal, no doubt these accounts are getting banned but likely not charged (who knows what jurisdiction the advertiser is even in) clearly they're able to pass the current bar.

This is conjecture but I'd bet that in all likelihood someone at some meeting at Meta has floated the idea of moving impedance to the signup to try to limit this phenomena. Maybe more verification, or more hoops. And likely Meta crunched the numbers and was looking at billions of dollars lost in legit ad traffic and that just doesn't fly for a publicly traded company.

I'd guess that the folks at Meta would absolutely love CSAM to stop showing up, It's simply too dangerous, likely costs more than it earns, makes them look bad when it shows up, and alienates users who don't want to see it.

But those negative effects are probably all quantifiable to some dollar value that's less than the alternatives.

Executives don't get bonuses for altruism or ethical behavior. They get bonuses for making more money than they did last quarter. real fix is changing the game but that's a lot harder than it seems and can have many unintended consequences of it's own.

curious why not integrate actual version control? Git or Fossil would seem ideal for managing changes in a book but the audience likely does not have the technical acumen to work with them easily. A client that can provide a strong abstraction but utilize the tool underneath would seem to have a lot of value-add.

Works on humans too. Garbage in garbage out.

First of all I don't think Meta is a good or honest company, but that doesn't mean they're not trying at all. In reality even if they were good and honest I genuinely believe their task is probably nearly impossible. Humans just suck too much. It certainly doesn't help that Meta has at least some perverse incentive to be permissive with ads, as they get money from them. Though Meta execs aren't stupid. Inflammatory political ads are one thing, kiddie porn is another I suspect they lose more money over CSAM than all the ad revenue from them. I believe them when they say they removed over 36 million pieces of child sexual exploitation content last year.

Which is a number that I should let sink in. Basically more than one piece of CSAM shows up in their ad stream every second of every day all year. If their moderation techniques are 99% effective that results in 360,000 CSAM ads leaking to the public annually.

Another interesting thought is how much would it cost to actually get that rate to 100%? let's assume human moderators aren't paid much but are infallible. That all-in a single moderator employee costs $65K , that's salary, benefits, HR ratio, psychological therapy, equipment, the works, all factored in.

let's say on average you get about 1 ad review ~every 2 minutes, so bout 250 ads per-day per-employee factoring in some time off you actually get about 65,000 reviews a year. So bout $1 per review. So the question begs how many distinct ads enter their system each year? which I haven't found a stat for. But assuming only 1 in 1000 ads are CSAM, which still feels high but is probably a reasonable ratio. we get 36 billion ads needing review annually. At a dollar each 23% of their gross annual ad revenue each year would be just moderation overhead and they would need to hire over 550,000 moderators.

Evil or not I think they organically built a thing that operates at an unimaginable scale, that even moderation with failures in the parts-per-million would result in tens of thousands of unsavory or unscrupulous ads reaching users and they have neither the incentive nor the capacity to reach even that level of accuracy.

[1] https://copypastatext.com/wp-content/uploads/2024/02/Doritos...

https://www.courtlistener.com/docket/4510056/arnold-v-maxmin...

Of course the best CGI back in the day was the CGI you didn't even notice was there. same with AI. when it's bad (and these days it's often bad) you notice, but then you don't notice all the places that. well you don't notice. And you didn't notice because there was nothing to notice.

I think this blog is pretty spot on, it's coming, people will seek out AI novels for, well their novelty (at least for a while), some will demand provenance regardless of quality, and everyone whether they like it or not will be consuming it and in many cases be none the wiser. eventually it will just be how things are.

I think they've been slow rolling it, eventually it will be login only.

in some sense you could push science so far ahead, germ theory, the structure of the atom, the immune system, silicon chips. All these "broad stroke" discoveries that you already know about. But when it comes down to brass tacks what is it worth? Like I know how a thermos works, and why it works, but heck if I know how one would actually consistently manufacture an evacuated double walled container. sent far enough back you would need to invent the vacuum pump, machines, maybe even steel itself just to produce a simple thermos.

makes you realize for all your knowledge you really never know absolutely everything, and that even simple things rely on chains of discovery, manufacturing and a whole lot of little details that we just don't think about ever.

I'm also not sure where the line between selective breeding and "evolution" , I wouldn't consider a Pomeranian an "evolution" of a wolf (though I guess it is a different species). Is any persistent genetic change "evolution"?

As for humans in general we've effectively eliminated most sources of selection, survivability really isn't an issue anymore, and fecundity is largely random, pretty much anyone can have kids, or choose not to have kids. The strongest thing we're probably selecting for right now is Mormonism and even they aren't making a dent in evolution.

man what a cop out, I was hoping they would use a text editor to build the gif. that would be way harder.

If you assume the cost of a thing can serve as a rough proxy for the energy inputs of the thing then until you meet RoI you're likely negative (environmentally speaking)

casual if the RoI for these is 4-8 years. and some stats[1] show most renters don't stay in one place that long. Given that the majority of people move at least once inside the RoI period, and may not move to somewhere with a balcony that can set up the solar it's possible that more of these system will end up in a landfill than actually help the environment.

which isn't to say I'm against deregulation here. I don't see any real reason to prevent people from being allowed to set them up so long as they're safe. and certainly some people will get great use out of them. but I also don't think they're as categorically great as they're made out to be.

[1] https://www.redfin.com/news/renters-staying-put-longer/

I suspect MUDs may have a bit of a resurgence soon, I've seen more and more interest in them as of late simply because they seem like an excellent pairing with LLMs which easily consume text. Though I think for your needs someone would need to work on the barrier to entry to getting on a MUD as well as the modern UI/UX expected today.

I don't think this tech is quite "full length movie" watchable yet (maybe other uses), but seems only a matter of time where it would be "good enough" for common use.

dunno, I've seen worse and some of the scenes if you asked me if it was AI or not I probably couldn't tell you. most egregious thing I saw was at the 27 second mark where the girl is inside the table.

At the current level of tech it feels like it would be a pretty dull full length Anime, and hard to squeeze all the bugs out but for something shorter like a commercial or FMVs for like a video game seems good enough.

Sad it's not available right now because talking about it puts a target on them for Nintendo to shut them down even if this effort is technically legal (not sure it is).

I wonder if this type of thing could cause a new age of game preservation, some games are just finicky and won't run on modern machines and throwing AI to port a code-base may allow some older gems to work properly again.

When does a feature that simply makes your product easier to use cross into a territory that it's illegal?

What about media previews, in a platform like reddit if you can preview or expand media directly from the main list is that an “addictive feature” or just convenient design?

also interested in the larger economy, if you create a plugin that restores or adds infinite scroll to a website could you be liable for providing illegal UX?

EDIT: to clarify I'm not really griping on infinite scroll in particular, more the difficulty in regulating postitive UI/UX. Dark patterns are relatively easy to identify. If the unsubscribe button is hidden behind 3 screens and is in 3 point font that's pretty clearly a dark pattern. This seeks to regulate the opposite basically "your platform is, too easy, too good at serving your content"

This is much harder to categorize and has fuzzier boundaries. For example imagine if it applied to remembering your login for more than 24 hours. Certainly people would use your service less if they had to continuously authenticate. Are long-lived sessions good UX or encouraging "addictive" behavior?

What were your engagement numbers between A and C? I would expect similar numbers between A and C maybe more with C since AI stuff can be off-putting to some users.

sounds like A and C are entirely different UX so maybe people just didn't like the results of A compared to C.

Unsurprising that later engagement with C was higher since signup would be a deliberate response to seeing the product instead of "because I had to".

Also unsurprising that B has the most raw signups because looks like with that there's no meaningful way to engage with the product without doing so, given the low impedance of leveraging Oauth (lots of people are signed in all the time anyway) for some users signups are basically zero-care, would be interesting if the ratio of email signups to Oauth on option C were different than B since C is a deliberate choice.

- Your AI code is probably as good as your human user's skills and attention:

Seems pretty likely they were shying from style guides because the AI use was itself sloppy. I'm not a big AI user but if I just ask a chat model for some code it will give it to me, then I have to go over it and fix the code for our check-style. Without deliberately aligning all your AI use to enforce it style is hard to maintain. Existing lack of style was a technical debt they didn't want to pay.

- The most popular language is probably the best language:

I suspect the amount of training data on Rust is several orders of magnitude greater than Zig. I suspect they could have gotten there simply asking the LLM to rewrite the Zig thing in Zig, however I'm betting LLMs write much better rust code so why not take the opportunity to move to a language that will get all-around better results.

- Technical bankruptcy is more favorable than paying technical debt:

These migrations we're seeing seem to point to a future where it's simply easier to burn a code-base to the ground, take the test suite and re-implement rather than actually pay large amounts of technical debt. I think this is a pattern that will come up more often with heavily AI written codebases that become untenably noodley, or brittle.

In an honest administration I would figure they would strike the requirement then go forward to answer questions like those to see if there's a need to form regulation around it. And an honest administration could probably bring forth those types of nuances to an informed and level-headed public and be accepted.

We do not have an honest administration, nor an informed or a level-headed public. His firing was likely both retaliatory for making them look bad when the information was released but also "we know this guy's not going to play ball and cause problems for us later. get rid of him"

1) Sony ending physical disks

2) potential high price of next gen.

1 is hilarious, even games that come on physical disks are already mostly useless without internet. I had a ps4 and bought "Last of Us" and the game is literally unbeatable without an internet connection. It has a game breaking bug that isn't patched on disk. Many other games are crap or broken without patches. Also jumping to PC where there hasn't been physical media in ages and 99% of games have DRM is just out of the frying pan into the fire. based on the market share for GOG I can assure you only a small fraction of gamers actually care about "owning" their games.

2 is kinda stupid, they mentioned a price point of $1,000, not sure you can build a next-gen-console comparable PC for that price. My current computer's GPU cost more than that by itself and anything but a pittance of RAM will too.

I mean go for it, more PC gamers the better (it's my chosen platform) but if you weren't already on board years ago not sure anything has really changed.

I wonder if enough of these systems reach return to cover all the ones that don't for some reason or another. Looks like ROI might be 3-7 years, at the high end seems probable people would maybe move and not be able to use them and never really get their money's worth out of it.