@silverwizard It wouldn't be the first buzzword disconnected from user wants, but I have been feeling pretty isolated railing against both the low inherent quality of LLM chatbots output and their massive externalities.
@silverwizard I must admit that I haven't read much AI criticism that I don't also share myself. The working theory I outlined earlier is based on a single poll that showed 40% of the responders were considering AI features at least somewhat negatively in new products.
When I learned about the poll, it went against my preconceived notion that the general public was more receptive to AI features, which would have partially explained the rush to implement it into anything and everything.
@silverwizard Oh I know I and the people I surrounded myself with don't want this privacy invasion. Beyond that, it becomes unclear to me that it would be the main factor of rejection. My main guess is there is a general disconnect between the data harvesting for AI purposes and the AI tools themselves. Meaning that people may be receptive to the AI tools and at the same time reject high-profile cases of data harvesting like the Reddit debacle, Twitter's change of terms of service, Recall, etc...
This would also mean that the general contempt for the latest AI-everywhere products wouldn't be based on the underlying data harvesting, but rather because they are disappointing products by themselves, regardless of their externalities you and I strongly care about.
@silverwizard I'm not sure about that. I feel like the general problem with AI is more that the promised convenience is disappointing rather than the privacy trade-off. Connected objects offer a tangible convenience (controlling apppliances from your phone) with a similar privacy trade-off and they don't get nearly as much hate as AI-powered anything.
@montag Das ist der Prozess, der alte Daten löscht. Ihn auszuschalten bedeutet, dass die Datenbank immer weiter wachsen wird. Ich habe bei mir dbclean-expire-limit auf 100000 gesetzt, weil ich beobachtet hatte, dass mein System sehr lange mit dem Löschen beschäftigt war. Ich bin mir nicht ganz sicher, ob jeder Datenbankserver so große Ergebnismengen beherrscht, deswegen ist das Vorgabelimit so klein.
Wenn Du den Wert auf '100000' setzt, wirst Du wohl zu Anfang noch nicht so sehr einen Unterschied bemerken, aber wenn Du Dir die Anzahl der Datensätze in der item-uri merkst, solltest Du feststellen, dass die Menge dann merklich runtergeht, bis sie ein Level erreicht hat, der dann dazu führt, dass Dein System diese Query nur einmal am Tag ausführen muss.
@liaizon I just stumbled across iconography.fediverse.info/ Do you have a downloadable archive for that and possibly a list of the respective licenses? I guess I would like to use these graphics for Friendica as a replacement for Fork Awesome.
@silverwizard This is disheartening as CSP are supposed to address exactly that. Either it's malfunctioning or is too brittle to be used effectively, and neither are a good look.