Post BABiSDyXMALlWp5Kgy by mousey@mastodon.seattlematrix.org
(DIR) More posts by mousey@mastodon.seattlematrix.org
(DIR) Post #BABgzWUJLg5GakhIPY by libreleah@mas.to
0 likes, 0 repeats
i recently wrote that I become anti-AI, having previously been neutral towards it - critical in some aspects, positive in others.when you use AI, you stop thinking. you don't reason about anything. google search really annoys me, because it's so seductive when the bot tells you what you want, regardless of whether it's true... and you're enticed to *not* click links anymoreanyway, i don't want to further articulate or intellectualise, but OpenBSD pretty much predicted this 17 years ago:https://www.openbsd.org/lyrics.html#46OpenBSD predicted chatgpt.This is really insidious. Even really smart people get sucked into it. My friend is one such person, but lots of really smart people I know seem to be taken by it.None of us are immune. When you have a sycophant whispering in your ear and tell you all the answers (whether they're right or not, more often they are wrong), you stop thinking.It's called: deskillingI say think for yourself. It's OK even to be wrong, or slow. Think for yourself!
(DIR) Post #BABhf3SZfFiMj2glhg by libreleah@mas.to
0 likes, 0 repeats
anyway, as a result of this policy change, i now actually *use* the built-in no-AI duckduckgo search on librewolf, instead of typing "google.com"not that duckduckgo is any good either. i mean, it's crap actually, i should set up searx or something, so as to continue using google (i assume searx can do google)i remember the world before AI. some people have called me a genius in the past, not because i know everything, but because of how quickly i learn everything. for example, i learned how to solder in a month, and got really good at it in a month. the photo shows how i soldered after a month of practise (this one several years after, but i did stuff like this after only a month)in 2013, i had no idea how a package manager worked, then accidentally reinvented package management when i started the libreboot project. libreboot is a package manager for building coreboot images. after a year, i thought: hang on, this is just like debian apt-get!i'm glad we didn't have AI in 2013
(DIR) Post #BABiSDyXMALlWp5Kgy by mousey@mastodon.seattlematrix.org
0 likes, 0 repeats
@libreleah "deskilling", thanks. This is a word I need. Your meme has successfully replicated, congratulations. I will be using this word like a bludgeon anytime I am asked to employ an LLM.
(DIR) Post #BABimUFIZc3DmWaSC8 by libreleah@mas.to
0 likes, 0 repeats
chatgpt didn't teach me how to do this. burning my thumb did. yes, i learned how to avoid burning my thumb.chatgpt can't do shit
(DIR) Post #BABjVHciLjpHfJtrjk by Razemix@mamutovo.cz
0 likes, 0 repeats
@libreleah Have you tried Startpage.com? AFAIK, it takes results from Google (and probably other sources too).
(DIR) Post #BABpdv7Vz7ZoCioBbk by libreleah@mas.to
0 likes, 0 repeats
speaking of librewolf. i recently also did this:lwmk: https://codeberg.org/libreboot/lbmk/src/branch/lwmk_wipLibreWolf MaKei'm much further ahead locally. i don't want to publish something close to completion so this is from an older branch i have locally.guess how much AI i used? none! yes, i'm forking the libreboot build system into a new project that builds librewolf, and i will propose it to librewolf when ready. libreboot's build system is highly generalised and configurable, and does a lot of things librewolf's build system currently doesn't, such as reliably checking revisions and checksums of sources that it downloads. i started doing it when i noticed that librewolf source.git lacks integrity checking on a bunch of things it downloads, such as... firefox tarballs.because it's lbmk-based, i don't use tarballs. i use git repos. this is bigger on disk usage, but more flexible for lbmk's multi-tree design. fork is: lwmkand here is my openbsd port, also created without AI:https://cvsweb.openbsd.org/ports/www/librewolf
(DIR) Post #BABqESe6oYTKbiThNA by thoralf@gruene.social
0 likes, 0 repeats
@libreleah This is an interesting take.For me, using LLMs does exactly the opposite. I do think a lot more ahead and plan much more carefully.Example: I had an automation for Homeassistant that basically deals with charging/discharging the battery, etc.With the help of LLMs I'm currently rebuilding that - but in a much more sophisticated way. I'm forced to think well ahead and build lots of scenarios to get a proper result.If I had done that by hand, it would have been trial and error.