In which I attempt to have a nuanced discussion about #AI and #LLM, which I will probably regret. But whatever, I don't think that black-and-white thinking tends to be useful. So grab a cup of your beverage of choice and settle in for a reasoned conversation. https://changelog.complete.org/archives/44740-ai-in-debian-the-vote-proposals-and-nuance
John Goerzen
mastodon 4.7.3Hacker, dad, pilot, amateur radio operator, activist, guy that is susceptible to new hobbies. Former president of Software in the Public Interest.
I live miles from the nearest paved road in #Kansas.
Interests: #rust #debian #linux #pilot #flying #hamradio #emacs #orgmode #kansas #floss #kansas #raspberrypi #programming #parenting #retrocomputing
SRE at Google. I do not speak for my employer; views expressed here are my own.
Long before #LLMs, two of the biggest uses for #AI were (and still are):
To show us ads we don't want to see (on the web)
To block ads we don't want to see (in email)
Funny how so many uses of AI are to make things we don't want.
Maybe in the future we'll burn a bunch of gigawatts with the ad-blocking AIs fighting the ad-injecting AIs, all to capture the attention of humans.
Or, who am I kidding, they'll probably be fighting to capture the attention of agentic AIs.
In which I try to look at #AI with a balanced frame of mind: https://changelog.complete.org/archives/42503-artificial-intelligence-shades-of-gray
I also discuss building a LLM email classifier using solar-powered local models, which a useful degree of success.
I'm neither a cheerleader nor a doomsayer. More a tinkerer.