someone's avatar
someone
npub1nlk8...jm9c
someone's avatar
someone 1 year ago
how to ruin a model v3: hermes3 the only fine tuning that was slightly moving in the right direction was dolphin from cognitive computations, months ago. i need to check what they are up to again.
someone's avatar
someone 1 year ago
since truth is generating from One and it doesn't change much, i.e. truth is ancient truth most of the time, the seekers of truth almost have to plagiarize each other all the time. people speaking about truth parrot each other. thats why building an llm based on it is a trivial task. each addition confirms the other. the numbers don't battle but just support each other. lie pumpers can get much more creative and nuanced and they should. lie producers should constantly distract and spend energy. truth is much more energy efficient. you can sit and shut down your mind and download lots of it. liars thru llms should spend resources to cover every kind of question that can be answered that will lead to a lie, in order to succeed.
someone's avatar
someone 1 year ago
awesome channel for fine tuning LLMs
someone's avatar
someone 1 year ago
a platform where individuals make their own ai and then other people combining decentralized truth into a synergistic thing could be a great option for battling centralized truth. the marketplace on that platform could allow 'ideas that work longer' to outlive others, and this eventually allow ideas that work longest to be accepted as "generally true". and this could potentially align the people's version of ai and this model could end up in brains for friendly robots. future is bright.
someone's avatar
someone 1 year ago
wow the replyguy halvened DAU! i think my relays handled him fine? i am checking the globals on coracle and did not see many replyguys? how was and is your experience with my relays?
someone's avatar
someone 1 year ago
lets see truthers hate ai bitcoiners hate ai nostr is uninterested doing a truthful ai is going to be very popular! ๐Ÿ˜…
someone's avatar
someone 1 year ago
blue text comes from a "smart" model. pink is from ostrich. you can compare a mainstream AI and what i am building.. many more examples i can send in DM if you ask. enjoy
โ†‘