Bodil

The Rolling Stone article doing the rounds has all the horror stories I've been afraid were going to happen in the interaction between LLMs and people without the ability to critically analyse the limitations of LLMs, except the scariest one of all: what happens when politicians start uncritically employing LLMs to make policy decisions.

We're not prepared for this as a society, and we need to start having these conversations right now before we destroy ourselves. LLMs may seem like cool toys or tools to you with your fancy critical thinking abilities, but a great many people are not equipped to deal with a machine that spews out platitudes on command so statistically credible they can't tell there's no intelligence behind them.

And if you think OpenAI and the likes are going to regulate this for us instead of continuing to rake in the cash, we're all the more doomed.

https://www.rollingstone.com/culture/culture-features/ai-spiritual-delusions-destroying-human-relationships-1235330175/

My Mastodon client is close enough to basic feature completeness now that I'm using it as my daily doomscroller, but it still doesn't handle polls attached to posts at all, and I'm rather coming around to the idea that maybe this is actually a feature.

jfc, I'm using the API docs published on Mastodon Dot Social to implement my Mastodon client, and I test on all sorts of things: Glitch, Akkoma, Catodon, I even set up a GoToSocial server to test on, and they all work exactly as the API docs on Mastodon Dot Social says they should. The single instance that ever gives me trouble in a diverging-from-the-docs-on-Mastodon-Dot-Social-in-unexpected-ways kind of way is, you guessed it, Mastodon Dot bloody Social.

I'm so tired, I think I'll go play video games for a bit now.