@jonny hard oof. I had signed up to the queue for DALL-E but actually I’m a lapsed amateur illustrator so perhaps I’ll just write down all those “wouldn’t this be hilarious” ideas as prompts for myself and get back in it or (shock horror) actually pay an artist.
It’s easy to get caught up in the hype and possibility of the thing but if the cost is artisans then fuck that.
@grimmware yeah, keeping the weights proprietary and charging for access is a completely different business model -- but one we shouldn't be surprised about. I think any company having the compute and expertise to make these kinds of models that are effectively magic is extremely dangerous. this is just one that we *know* about, not to be a conspiracy theorist but the possibility of sophisticated deepfake models we *dont* know about it always there.
@jonny I’m amazed how many companies are able to jump on the bandwagon of displaying their modern values (e.g. “no means no and any further pursuit is harassment”) and then immediately make a terrible consent faux pas in literally any other place.
GitHub Copilot, DALL-E, any website or app which has a pop up with the options “yes” and “as me again later”.
And by amazed I do of course mean “wholly unsurprised at how it all unfolded with a tedious inevitability.”
@jonny I’m really pleased you mentioned this wrt DALL-E because I knew intellectually that this was an issue with AI but got totally caught up in what could be done easily with DALL-E and somehow didn’t make the mental translation. You saved me from compromising my values there <3
yno I feel like there's space for poking at even evil tools without celebrating them. like using dall-e art to inject this into the otherwise somewhat monolithic "whoa this is sick" conversation would be useful praxis I think. but ya the broader problem of these companies and their business models and unaccountable/uninspectable magic computer tools remains.