TIL that @grok seems updated every 24 hours or so. Because in an answer it brought up a very niche reference related to a post of mine from about 24h ago. So post data either does indeed feed back within 24h or it's lying about profile search on the fly
@erikphoel is the failure to find more recent polymaths mainly a function of increased specialization or widening of areas of possible activity? So that it is harder to construct a valid yardstick what counts as polymath. Is being an expert video game player part of today's polymathyness?
@gummibear737 for me the most revealing snippets are often the ones were the nonsense comes out so matter of factly that you know he didn't make up a bold lie in that moment but just has a funny idea of the world out there https://t.co/0nkCL1Z4Fx
@teortaxesTex yes you are right that what i try to understand here is the typical quality of local-global spatial structure of the image itself. So that probably less to do with prompt processing llm part but rather the diffusion model generating the image
One thing i'd like to understand: this give away ai style with 'detail greeble at scale impressing through spatial coherence to large scale shapes' If this is a tendency emerging out of llm architecture, or is mainly result of popular taste in original training data?