avatar
Miriam Posner @miriamposner.com

Who’s dismissive of the tech? I teach how LLMs work from start to finish. I just don’t pretend that elementary school kids need to learn how to fabulate memes.

aug 30, 2025, 5:04 pm • 4 0

Replies

avatar
Luis Villa @lu.is

“It is extremely easy to use a chatbot” is… pretty dismissive? To be generous, I suppose this is dismissive of the user and not the tool, in much the same way “it is extremely easy to use a saw” is dismissive of the skill of the carpenter rather than the tool. So 🤷🏽 Anyway, have a fine morning!

aug 30, 2025, 5:09 pm • 0 0 • view
avatar
graeme @g--0.bsky.social

the objective of a chatbot is that it SHOULD BE extremely easy to use. It’s not the users fault if a product doesn’t work properly and there should be no expectation that the user base simply learn to use workarounds vs demanding the flaws be fixed

aug 30, 2025, 5:18 pm • 0 0 • view
avatar
graeme @g--0.bsky.social

if you’re riding a horse, the animal will do all sorts of stuff you don’t want it to do and you’ll have to use various tricks and techniques to deal with that. This should not apply to human created tools designed for use by humans!

aug 30, 2025, 5:19 pm • 0 0 • view
avatar
graeme @g--0.bsky.social

we used to call these problems “bugs” and software companies would restrict them to a small user group called “beta testers” until they were resolved.

aug 30, 2025, 5:20 pm • 1 0 • view
avatar
graeme @g--0.bsky.social

the fact that the output of a computer program is non-deterministic doesn’t mean humans should be expected to learn anything extra, it just means the program sucks and shouldn’t have been promoted to production

aug 30, 2025, 5:21 pm • 1 0 • view
avatar
Luis Villa @lu.is

I don’t entirely disagree, and this is part of why the most interesting parts of the industry are moving away from chat as the dominant model.

aug 30, 2025, 5:29 pm • 0 0 • view
avatar
Luis Villa @lu.is

But also I would love to see some historically informed discussion of this, because “it should be perfectly predictable and reliable” is a very recent mode of thinking about machines? The overwhelming history of “tech” is that it’s a finicky beast that requires attention and rewards tinkering.

aug 30, 2025, 5:29 pm • 0 0 • view
avatar
Luis Villa @lu.is

Think (most creatively, and not coincidentally most ultimately destructively) hot rodding as precursor of modern car culture, or the intensely craft/techne-driven relationship of early printers to the author and the printed book.

aug 30, 2025, 5:29 pm • 0 0 • view
avatar
graeme @g--0.bsky.social

nobody has access to the internals of a chatbot like the owner of a hot rod has access to what’s under the hood. Users of LLMs are in the drivers seat and the hood doesn’t open

aug 30, 2025, 5:32 pm • 1 0 • view
avatar
Luis Villa @lu.is

Hi! I am such a FOSS guy that I’m the only person (other than Lessig?) to have served on the boards of both the Open Source Initiative and Creative Commons. I’ve spoken repeatedly on ethics at FSF’s annual conf. I can talk tinkering with software, and its moral and practical nuances, all day long.

aug 30, 2025, 5:42 pm • 0 0 • view
avatar
graeme @g--0.bsky.social

that requirement is only true if the tool is being used for things where the result matters, yes

aug 30, 2025, 5:31 pm • 0 0 • view
avatar
graeme @g--0.bsky.social

“tinkering” is part of programming yes - specifically it’s part of the development cycle, the part before promotion to prod

aug 30, 2025, 5:31 pm • 0 0 • view