Ancestors

Written by Kye Fox on 2025-01-31 at 22:29

I don't think the anti-AI crowd hates good uses like climate modelling, but I do think they're easy marks for people who want to use their ignorance on the subject to attack the sciences because they don't realize how much AI/ML is in their world already b/c they think it means generative stuff.

=> More informations about this toot | More toots from Kye@tech.lgbt

Toot

Written by Kye Fox on 2025-01-31 at 22:32

I spent time learning what I could about LLMs, their predecessors, and all the stuff that's been added to them so I could converse intelligibly on the topic. This led to me finding ways to use it responsibly. Knowing stuff rocks.

=> More informations about this toot | More toots from Kye@tech.lgbt

Descendants

Written by Kye Fox on 2025-01-31 at 22:35

Actual, real example: I used ChatGPT to collect the top 10 episodes of Stargate and reviews for a project. This took one little prompt for something that would have taken far more energy to do with a traditional search engine, not to mention hours of my life. Verifying what I'm unsure of is quick.

=> More informations about this toot | More toots from Kye@tech.lgbt

Written by Kye Fox on 2025-01-31 at 22:42

My general view is that if the thing would have normally taken a lot of time and energy with traditional approaches, dropping in to ChatGPT is at least ethically matched to using a search engine. For everything else, a collection of local models on nuclear/hydro/gas.

=> More informations about this toot | More toots from Kye@tech.lgbt

Written by Kye Fox on 2025-01-31 at 22:44

The copyright question is kind of moot in 2025. If any of the training data is in there despite the processes used to keep it out, it's too sparse to come out unless you go out of your way to generate it or use a too-simple or obsolete model.

=> More informations about this toot | More toots from Kye@tech.lgbt

Written by Kye Fox on 2025-01-31 at 22:49

And, more to the point, if you're using it to make stuff you don't plan to treat like pre-writing and use it as-is, you've missed the value of the thing and you're limiting your creative growth by letting a mindless machine stand in for your brain.

=> More informations about this toot | More toots from Kye@tech.lgbt

Written by Freya (it/its)𒀭𒈹𒍠𒊩 on 2025-01-31 at 23:29

@Kye our current LLM if we lack the spoons to do text analysis is Deepseek-R1 running on a deliberately power-limited box (12W max, 12V 1A) running Solaris. Is it slow? Sure. But can we leave it to do its thing? Also yes

=> More informations about this toot | More toots from freya@chaosfem.tw