Notices by Erin 💽✨ (erincandescent@akko.erincandescent.net)

Many social networks have an artform. On Twitter, it was about brevity; and on Cohost, it was the CSS crime

And on Fedi, it’s composing the post with the most unhinged but accurate content warning.

Novel ways to cause an outage: go compiler ate all the ram. oom killer looked around to pick a process to kill. oom killer decided that the process it should kill was bird.

@soatok @ariarhythmic @notthatdelta @FurryBeta

Housekeeping gets paid in tips

What the Fuck North America.

@trashheap @cwebber @fay I think “the model sometimes regurgitates verbatim fragments from the Linux kernel” or such is not a massive problem. You can screen for this and get the model to try again.

@cwebber @fay Right. If you are OpenAI/Anthropic/XAI/et al, the ideal outcome of everything for you is

• Models are definitively derivative works without fair use exemption, because then all models will be closed and everyone will be forced to pay you to use theirs.
• Outputs can be (but are not necessarily) derivative works, because you can do deals with the people most likely to sue and with the information brokers you purchase training data from

I think we all prefer a world where, if you want to use a model (LLM or otherwise; I’m sure we’ll see better systems than LLMs evolve over time), you have the choice of running it on your own hardware, where all your inputs and outputs can be private and not exposed to a handful of massive corporations.

@fay @cwebber I think most of the opposition I’ve seen will only result in net negatives.

I have no issues with the idea that the output of a model can be a derivative work of the training data. On the other hand, if a model is a derivative work of the training data (and this is not fair use or local legal equiv.), the only models will be closed models, and that’s a significantly worse world.

@cwebber 🤖🤖🤖 (i’m not an agent but i want a surprise)

@ledoian @dalias it’s Linux, you’re root, you can just change the code to simply lie or do whatever you want, you have that capability.

Unless someone legislates that you can no longer actually control your own computer (and yes, people are trying to do that), or systems are legislated to collect some kind of proof but that’s a completely different legislative problem

@dalias @robryk Can you tell me which industry is pushing for social media bans for under (insert jurisdiction dependent age here)s? Because it surely can’t be the social media industry which stands only to lose users from this and I can’t see anyone else who is at all affected by this

@emma @dalias hey i’m all for these things but also think maybe we should do things to stop young children from accidentally wandering into pornography (especially but not limited to, to use an example, things like CNC scenes absent the context to understand things like pre-negotiated consent) or violent movies

@dalias i’m not sure what your actual argument is here.

Is it

• “We shouldn’t provide parental controls because instead of using them responsibly to give children access to developmentally appropriate things they’ll abuse them to restrict what they can see”, or
• “We shouldn’t provide parental controls because instead it should be the responsibility of those same abusive parents to watch over what their children are doing”, or
• “We shouldn’t provide a system level age restrictions API because the parent might provide an accurate age to the relatively trustworthy computer (which will inevitably result in them providing it to relatively untrustwothy apps instead)”, or
• “There is absolutley nothing harmful to children available on the internet and we should just provide them 100% unrestricted unmonitored access without age gates of any kind”?

@dalias A date of birth field in a user information record is an abuse mechanism?

@dalias abusive parents will surveil and control their children whatever you do. Honestly if some of these parents decide to leave things up to the government (which is on average midly conservative) instead of themselves (which is quite often incredibly conservative) it might even be a net win

Everyone seemingly getting mad about systemd adding a completely optional date of birth field to user records that is, in reality, only ever going to be filled in on the machines of children administered by parents who want such restrictions enforced, perhaps on machines administered by schools, or by people who want their computer to wish them a happy birthday.

@cwebber i’m not sure how, claude is running inside a read only github actions sandbox.

@cwebber FWIW I think they’ve been becoming less sycophantic over time. One of the overwhelming complaints I’ve seen over the past year is about the incessent plattitudes, and the model vendors are definitely working out how to tone that down, to the point that “you’re absolutely right” already starts to feel like a dated reference.

Of course they continue to be servile, they’re just becoming less blatanly sycophantic.

@cwebber @promovicz @laurenshof @evan I think one of the problems we’ve had in general is that signing things is a bit of a nightmare. Not just from a non-repudiation perspective (ActivityPub is pretty crap at this - though workable workarounds sort of exist.. - but I doubt ATProto is much better) but from a revocation and propagation of outdated/deleted information perspective.

Why do we not sign things? Because we don’t have a revocation story and also because indirect relaying gives up all sorts of control. Why is ATProto a bit more flexible here? Because they gave up that control to begin with.

If the signatures had expiries (which as far as I remember, they don’t!) you could imagine a world where when you click the boost button on my post, you ask my server for a copy of the post that’s signed and carries a short lived signature and then you would relay the post alongside that signature; but then it turns out that one of your followers is on a server that I blocked and now my post is there and, as a general rule, the Fediverse has decided that this is unacceptable (despite being unenforcible in general!), mostly as a consequence of the fact that we don’t have any form of 3rd-party-enforcible reply controls (I wish we had that, maybe it’ll come as an evolution of Mastodon’s quote controls…)

(And yes, LD Signatures suck, but all signature formats suck in some way or another and signatures are a primitive that it really sucks to build things around. But that’s a whole separate discussion!)

@dalias @cr1901 yes. well it doesn’t really retransmit, the packet will get dropped but it’ll now know the path mtu so it can do sender-side fragmentation.