Airbus open sourced their new cockpit font. Make it the default for all your embedded projects, because it's REALLY good and has real testing in difficult environments!
⛈️ Information ⛈️
Uninteresting. Genderfluid. No interest in performing for you. Was KirinDave on 🐦. Still KirinDave on YouTube.
$megacorp SRE. ML/Art/Software.
Not a fan of this capitalism. I am a fan of biocosmism, anarchism, distributed and robust systems, and economics when decoupled from capitalists and oligarchy.
ML models *must* be fully open sourced, including the training data! We deserve the right to audit.
Turns out a human centric, low bandwidth, ad free internet was possible all along. It just had to cut costs for tech vendors to make them amenable to permitting it.
https://thenewstack.io/cloudflares-markdown-for-agents-automatically-make-websites-more-aifriendly/
Ugh, all my EndeavorsOS machines came up with major issues and the website has been down for days. I'm literally torrenting the backup imagine and validating it with way back machine :(
I understand a lot of the anxieties and legal concerns that LLMs bring up, but I'm still baffled by folks who say they're useless. As documentation searching tools, they're simply unprecedented.
It's a bummer to see Doctorow believe blatant misinformation in his latest article. It's one thing to not like LLMs, it's another to push widely discredited (and easy for LLM providers to Motte & Bailey) water pollution/depletion narratives.
Believing false things will only hurt any chance at true harm prevention.
When IS the appropriate time to say, "I told you so" about programming AI models obviously growing in capability?
Like do we wait another year or can we just do it now?
I guess I shouldn't be surprised but I'm kinda amazed Ed Zitron isn't more contrite/embarrassed about how wrong some of his claims and predictions have been.
So Doctorow has been using LLMs this whole time? Lol.
Copyparty is such a godsend of a piece of software. It fits into a really good space of "extremely easy to set up" and "extremely useful feature set" without being "overpolished" or "sacrificing functionality for the sake of an ideal user who thinks a button on a website is offputting."
Oh reading that article, Zitron also says that GPUs and TPUs are only useful for AI and that's crazy talk. They're the backbone of scientific computing and likely to increasingly displace aspects of the classic CPU architecture over time.
Having taken a lot of time to think about it, I think that the Bluesky model for PDS's is objectively superior as a way to structure and empower users directly. I think in this *specific* dimension, the Fediverse would do well to adopt that standard and move, and I doubt it'd even be difficult to do.