Valerie Aurora 🇺🇦100% agree with this post
| Blog (mostly idle) | https://www.curiousefficiency.org/ |
Alyssa Coghlan
@ancoghlan
- 300 Followers
- 232 Following
- 1.3K Posts
CPython core developer, Python platform engineer for an Australian retail bank, cognitive science dabbler, secular humanist, charitably mercenary cynical idealist :)
@JessTheUnstill It's definitely harder, since so many programming tools aren't internationalised, and hence report any error and warning messages in English rather than the user's native language. Even translating docs requires a big time investment, though - translating the error messages not only needs the time investment to maintain the translations, it adds more opportunities for secondary failures when trying to report errors.
Mario MunozRE: https://cosocial.ca/@timbray/116337461606384177
From the shared article:
"But folks annoyed by online animosity toward AI need to remember it's an extension of justified rage at the extraction class' grotesque enabling of the literal destruction of democracy. That rage is going to prove extremely helpful in the months and years to come, and deserves a wide berth."
@jacob @pythonbynight @ambv "Explainability" is the word generally used for this requirement when it comes to the use of machine learning systems in finance (loan approvals, that kind of thing), so if there's a better word, it hasn't been found yet.
@robryk @mjg59 Ah, I missed mentioning one of my functional requirements: the end result needs to be executable by normal people so they don't reach for the "Eh, I'll just let the robot impersonate me with full access to everything I have access to" hammer. I'm trying *not* to see this as an inevitably doomed exercise, but at the moment I don't even know how *I'd* do it (beyond the tediously painful to set up "lots of fine-grained keys in a sandbox" approach), let alone help someone else do it.
@mjg59 This question gets even more fun when you'd like to give read access to an agent to enterprise services you otherwise have read/write access to. State of the art seems to be "carefully set up narrowly scoped API tokens for the services that support that", because constrained delegations of authority just aren't part of the overall identity management model, they're only present inside specific applications. (I don't have an answer, just the same and related questions)
Huh, I appear to have inadvertently discovered a way to keep the owlbear mother *alive* in Baldur's Gate 3: I accidentally started the fight with her, and ran away rather than finishing it. Finally go to the goblin camp many long rests later, only to find the actual chicken rather than the cub. Go check the cave. Mother and cub both still there, both still hostile. #bg3
Ash K🏳️🌈 Brisbane Queer Tech meetup for April: come hang out with me and make some IRL friends at the Ship Inn. I figure most peeps on fedi are a little bit tech, so no issues there. Are you in? RSVP at https://luma.com/bs5smgmw?utm_source=fedi and pass it on :)
@kytta @carlton @hynek Yep, this. The idea of cooldowns is:
* default is to wait before picking up releases
* folks willing to do their own security testing pick them up eagerly
It's similar to being *able* to opt in to consuming prereleases for compatibility testing, while the default is to wait for the full release.
UX details in both servers and clients (like offering "release X of package Y" granularity when bypassing cooldown configs) is still very much a work in progress, though.
@hynek Fun one I ran into a while back: the dedicated torch index server doesn't publish upload times (or at least didn't at the time, I haven't checked lately), so if you turn on upload time based filtering in uv the non-pypi torch wheels get completely ignored.