Hm, I think there’s more to it than that. It is often the case that someone will do something and regret it without having necessarily gained any new information, which is in a sense irrational, right? If you didn’t gain information, you could have known to just not done it to begin with. Something about experiences, perhaps?
Yeah, I'm interested too. My guess for the reason, if not purely to eke out more performance, is so they can cleanly gather real-world data on this kind of usage.
Rolling average including vacant months as $0? And if that isn't smooth enough, add some smoothing factor, count vacant months as 10% of the last paid value, or maybe the first vacant month as 50% with further months decaying. Or some other fancy accounting that makes more sense than the current method.
1m^3, right? I can picture what you mean, but I'm not sure it works technically, since I think the splats for a given region are not actually bound to the region they represent. Like, for example, reflections work by having the reflection being physically behind the reflective surface. And they're all transparent, so it'd blend together.
Somewhat related, check out https://github.com/wordbots/wordbots-parser, a digital card game where you write the cards and the engine parses them to determine what they do. It's fun to mess around with.
You gotta zoom into the center to see the main chaos. And if you zoom in elsewhere, you can see all the generation differences from different versions of the game.
I thought "A decentralized music tracking and discovery platform" and the features list was plenty descriptive. I don't think you have to use Bluesky to use this, I think Bluesky is like the backend, and otherwise it's a Last.fm alternative?
Neat. I wonder if a allowing the models to inspect pixels or pixel regions, instead of fully relying on the VLM, would help at all. The spatial reasoning required might be too complex though. In general the VLM seems to be a limiting factor, so I wonder if there's some way to usefully augment it or sidestep limitations.
Like, instead of being in pseudo-MSpaint, pseudo-Photoshop with manipulable layers and bounding boxes. They struggle to add an outline to something previously drawn, but that's something that could be done programmatically. The limitations are obviously part of what makes this interesting, but different limitations could be interesting, too. Maybe additional complexity would just result in more uninteresting failures though, I don't know.
I noticed that the feedback/strengths/suggestions outputs are clearly also given the initial image's prompt. It could be useful to additionally have an output that's not given the prompt, so the LLM knows what the VLM sees without bias?
I think it's a mistake to view any politics as bolted on. I think it's unlikely some people were interested in "mindful and resilient and ecological use of computing" completely apolitically, with no other political or ideological background.
This principles page doesn't seem to have any irrelevant politics to me.
Great, what is being done to help that happen?