A lot later and miles away: Trying to step back a bit and watch ones own stream of consciousness dig through similar topics again. Hesitating to take the time drafting an image of ones inner map. Maybe outcome would be either scary or trivial. Not sure which might be better. Not likely to ask the model about that, for it's quite likely to come up with answers even less helpful by now. Watching, listening, smelling the village and its green vicinity slowly drift into sleep again instead. Have a calm night wherever you are.
Entlang des Weges: Der Wetterwechsel, der plötzliche feine Regen, die nassen Steine und Schuhe. Dann Büroküche, Büroflur, Schreibtisch, Kalender. Ein Modell, eine wie immer unbefriedigende Antwort, Schulterzucken und das Gefühl ausgeprägter Reibung an nur allzu vielen Stellen. Noch sammeln sich die Stunden.
After all: Letting pass a day of rain and sun gambling without either one really winning. Trying and struggling to brush off the greasy dust of the hours. Once again, involved into strong disputes with various inner voices, again without either one really winning or making a better point. The model doesn't know any better than echoing the loudest arguments, and maybe that's part of the very issue. (Still about to step out of the flow. Wondering whether that border of being tired enough to sleep has already been crossed again. Have a soft night wherever you are.)
Clouds, slightly more dense by now. But probably still not reliable enough to bear the weight of todays backup. Or: Of context and language and confusing terms for what seems the wrong things. Tempted to get the model involved into this kind of ramifications but then again there has been enough resources spent on nothing again. Like on any other day too.
Also, elsewhere, talking large language models, image generation and everything related to that:
Miyazaki himself has called AI “an insult to life itself,” while fans of his work happily and unwittingly feed the hungry AI machinery with opportunities for iterative practice, helpful feedback, and subtle signals for improving the technology along the way. All while completely disregarding the crooked practices that went into sourcing content for the models, and totally disrespecting the opinions and wishes of the creators of the original art upon which these cheap derivative imitations are fabricated.
Watching this happen is really frustrating. Watching artists see the time and talent and passion that they put into their work being so casually devalued by an increasingly effective array of smoke and mirrors (while simultaneously witnessing the harmful ecological side-effects, as some kind of demented cherry on top) is even more frustrating. And then for me, personally, being asked to weigh in on this, to take a strong position and ban the use of this technology within the omg.lol community, well, that’s extra frustrating. But it’s happened a few times in the past 24 hours, and so here I am talking about it now.
I'm having quite mixed opinions on most of the AI and LLM topic going on out there for a bunch of reasons, somewhere in between the general attempt to be basically open in regards to benefits and hazards, advantages and negative side-effects of whichever tool comes up next - and bad gut feelings on many different levels, including distribution of power due to model control, hardware costs, energy consumption, infrastructure spent on such tools, and including effects like these. For this very aspect and in this situation however, once again I completely side with Adam and his rather well-written article, as well as the conclusions he is about to make here. Worth reading.