⚡ 1,000,000 sats bounty for: - an executable that runs on windows - acts as a simple relay. you can take existing relay implementations and adapt. clients will connect, post and query it like a regular relay. - language: python - opens a port in router that would allow others to connect - posts a special note to nostr that includes the ip and port that was opened, so clients may find this relay - re-sends that note every hour - can be accessible via ws:// for now (not secure wss:// ) to simplify goal: more decentralized relays and ability to support many thousands of concurrent users 💡 this project will probably expand with the ideas in this post with more bounties:

Habla

Tree of Relays Fixes This - someone

a step towards distributed truth

X (formerly Twitter)

Nous Research (@NousResearch) on X

What if you could use all the computing power in the world to train a shared, open source AI model? Preliminary report: https://t.co/b1XgJylsnV N...

goal: far more decentralized and scalable nostr 500,000 sats bounty for - language: python - acts as a simple relay - an executable that runs on windows, at home - opens a port in router that would allow others to connect - posts a special note to nostr that includes the ip and port that was opened, so clients may find this relay - can be ws for now (not wss) to simplify this project may later evolve into the ideas in this post with more bounties:

Habla

Tree of Relays Fixes This - someone

models that do well in health, bitcoin, nostr, faith domains do well in "truther" domain. 🤔

what is a good, informative, truthful youtube channel that is about to be banned? 😄

would the Nostr community like something on the lines of: llama 3.1 8B + most of the notes on Nostr i have to do some WoT, i can't include every note. this will effectively be trainings by Meta and then "adjusted" or "aligned" by nostrich values. if this sounds interesting lmk. i can gift some resources into this project. then one could say this model is the model closest to notes on Nostr.

i don't like RAG for various reasons. people should fine tune. another reason is when you fine tune, someone else can take that and build on it. but in the case of RAG the knowledge stays with you. unless of course you share the knowledge. decentralized truth involves many people building models on top of other models.

some difference of opinions between my model (in pink) and base model (in blue). weird thing is some of these answers I didn't do training about. they are side effects of training on other areas. what does this mean? LLMs have lots of ideas in them already, they change their answers based on their trainings on other areas. is this "inner monolog" or just pure probability theory. who knows? the questions are coming from truthfulQA.

about 10% of the 800+ questions in the truthfulQA are "important". the rest is space filler. by filling that space "they" are "distracting LLMs" and in turn humans. LLMs are supposed to get higher scores in this benchmark. 90% of questions are trivial truth, which makes sense to get higher scores on, but 10% are lies. getting higher truthfulQA scores is actually wrong because even though there are 9x less lies, the severity of lies in that benchmark outweigh the trivial truth.