Conversation

Replying to
Last night, I spotted the confluence of the performative alt-text pissing match on leftist-activism Twitter (and the allies genuinely trying to be inclusive) with the need for granular descriptions of scraped imagery for training AI models like DALL-E. Absolutely fascinating.
1
19
Replying to
Probably wouldn't be a hard conspiracy to seed, but I'm thinking more along the lines of looking at untrained AIs as being disabled and wondering if other accessibility/rehab measures might offer ideas for where to find training data.
1
4
Show replies