🚲

dx@social.ridetrans.it

I toot for me.Utility bicycling, playing cards, and grumpy shitposts. Transphobes and fascists can get wrecked. The world is on fire and we should try to save some of it before there's nothing but ashes.lowercase toots are usually jokes 🚲 🚶‍♂️ 🚉 🤖 🧗‍♀️ 🎲, 𝋑𝋌

Posts

Post #3077132
It’s equal parts invigorating and depressing to watch Mamdani’s progress. It’s like, all those things they claimed “can’t be done” or “are too hard/impossible” actually are doable. Which means that all the other politicians are not doing them as a *choice*. Which I kind of knew already but Mamdani draws a line under it.
Post #2198911
Is there a way to invest in stuff like solar and enhanced geothermal on a broad level in Canada? Like, aside from my morals they just seem like good bets. I want something fairly broad but more targeted than your typical etf
Post #2198903
I would bet dollars to donuts that if you tried to erect a sign that said “Liberty and Justice For All” there would be an uproar about it being too woke, even though it’s from the Pledge of Allegiance My hometown had a whole small town drama about them putting up some “All Are Welcome” signs and the local fascists not liking that.
Post #2198902
ten years ago there was some interesting work on building deep classification models that not only made a classification but also made a confidence estimate (and not just taking the softmax score as a pseudo confidence). I wonder why something like that isn’t used for LLMs, maybe an auxiliary lighter weight LLM that’s been trained explicitly as a bullshit detector. A bit like LLM as a judge but more principled. (1/2)
Post #2198901
We put a small table next to our front door back in October for a bowl of Halloween candy and we haven’t moved the table because it’s actually super nice to have a table height surface to place things on while you’re unlocking the door
Post #1490752
It’s pretty funny that the AI is generating human comfort programming languages instead of machine code or at least something lower level than Python and typescript or whatever. I guess that’s where the training data is.
Post #1100793
This is my tiny hill I will die on.
Post #728992
If you ignore the ethical and moral issues around the training data used by “AI”, there’s still the environmental issues. If you ignore the environmental issues, there’s still the social and societal issues. If you ignore the social and societal issues, there’s still the financial issues. If you ignore the financial issues, there’s still the technical limitations. If you ignore the technical limitations…