asura
they / any and will move your infra to GCP for food
Broke Blaise's (@wittywebhandle.bsky.social) brain after they claimed batteries weren't a form of vertical integration but openAI was. 🤪
Tech, pets, games, ... Linux, python, cloud
Be queer, do things
- the state of ai chatbots on twitter is getting absurd. using an ai chatbot to fact check an ai chatbot
-
View full threadWell, Neuro’s creator Vedal apparently spends a massive amount of money on maintaining and upgrading Neuro, but I don’t think it’s a result of huge energy consumption.
- Most likely a lot of nvidia cards :P
- I found this comment from someone explaining the basics, which I kind of understand.
- Thanks, yeah that's what I thought. I was really fuzzy on which bot was which and for a second the commenter above had me thinking neurosama was moved to chatgpt or something Local models barely make a dent "resources" (my electric bill) over my PC on or playing games, neurosama would be the same
- Yeah, using grotesque resources for a stunt is sooo funny.
- What does neurosama run on?
- Things that are still MASSIVE markets despite no longer being the subject to remotely as much speculative investment that they once were: - tulips - orchids (basically Victorian tulip mania) - coins - Beanie Babies - baseball cards - Pokémon cards - high-end wine - domain names - NFTs
- you forgot - developers
- Arch - use mesa git Steamdeck - supposedly works at 30fps on latest??? Fedora - use mesa from xxmitsu sudo dnf copr enable xxmitsu/mesa-git Update to latest and go Doom: The Dark Ages
- We doomin' in linux Doom: The Dark Ages
- With regards to Hasan Abi, he specifically mentioned he already knew he was in a position of privilege and a public persona It was a gamble and a gamble more people in that position should take. Get a lawyer but win playing chicken with the "deals" admin because brown and trans people can't
- [This post could not be retrieved]
-
View full threadand we still havent worked out how these things scale with input data, and even with access to all copyrighted works these things are starting to show serious limitations This is inevitably going to lead to AIs training on AI created data, degrading outputs
- That's not inherent to the technology. That's inherent to the provider you're using - hint: maybe don't use grok :P Here's an example of an actual open LLM
