Analyst signüll argues OpenAI's Sora video model diverted critical compute and damaged enterprise trust

@signulll this is possible but it is a bit funny when you consider the tradeoff between making a funny video app vs the “safer, more ethical” choice of focusing entirely on making a potentially world ending superintelligence

@tszzl one is a known nuisance, the other unknown. tbd which is which.

@signulll Also the sexbot thing was a pr detour and brand degrading experience that compounded

@signulll

@signulll OpenAI really lost focus. Anthropic appears to have had more of it.

@signulll Karpathy is one of few for the long haul. But everyone and everything else seems like a Google style project. Be it stadia, reader or plus. Agility and even shorter attention spans. Andy Warhol’s 15 mins condensed to Tik Tok’s 15 secs & Hawk Tuah.

@signulll Possibly, but the research and implementation of Sora resulted in a lot of valuable learnings about world models, directly applicable IMO to their robotics program. Still probably less-than-optimal way to get there though.

@signulll Also the price of chaos

@signulll Yeah, distribution strategy leaks into research strategy. Once you ship a consumer surface, you inherit moderation and brand costs that have nothing to do with frontier progress. A flashy demo can be locally rational and globally expensive.