Peatbot
-
@Living_thing So glad to hear it helped you. It's not going anywhere!
-
@brad good to hear it
-
@brad I've noticed some activity on telegram with people trying to feed peatbot into a monetised LLM.
Is there any chance of having a peatbot TG channel so that others don't rip it off on that platform?
-
@Living_thing said in Peatbot:
people trying to feed peatbot into a monetised LLM
Sounds like a new and amusing depth of stupidity.
-
@ThinPicking Fr. it's this jewish guy trying to make a subscription based site where people have to pay with his crypto token to access their AI doctor.
-
@Living_thing said in Peatbot:
jewish guy
I don't think that's relevant but for the mention it might be our very own Sharko. In which case he may just be mentioning it for comparison. In his thread he described his own attempt using oAI's RAG. Which would be populated with the same or similar source material.
If it is that, let him cook I guess. If you build it and they don't come, you learn.
-
@ThinPicking Its a user who goes by Alexander who runs a crypto channel.
He is part of the APU meme coin team. -
@Living_thing said in Peatbot:
runs a crypto channel
Oh dear. Just the type to do the bad (stupid) thing.
-
@Living_thing Can you send me some details? Just want to make sure they don't scale it up because that would make it quite expensive for me to run.
-
@brad I found out that its a copy rather than the real thing.
It gives gpt like responses at times.
Its just a bot that was trained on ray peat content.
the token ( $arcdoc ) is dumping now. -
-
@ThinPicking XDDDD
-
@brad Hey there Brad. Do you have the source code anywhere? If not, can you please make the model behind it one from the Gemini series. They have a much larger context window, which could be helpful for many
-
@lobotomize-me Not exactly but this should give you a good start https://colab.research.google.com/drive/1Gj5UwTCdJFiJzS_TQ9m1pPkio9jFEWlV
-
PeatBot stopped giving responses... Is it a bug or a feature?
-
I’ve got a new health bot that will have a forever conversation with you — tell it everything you go through day by day. It will be your health buddy and help find patterns, make suggestions, etc.
Like your good doctor friend, this will never run out of memory. You can use it for years and years and it will only get smarter.
It will take the Ray Peat theories and use those, along with other legit “alt health” sources…
And it is fully private, with all memories stored on your device(s).
I will post an invite link when it is ready for beta testing. I think you will absolutely love it.
-
@Ecstatic_Hamster said in Peatbot:
It will take the Ray Peat theories and use those, along with other legit “alt health” sources…
Hammy has the grand unified theory and all he wants to talk about is a bot.
-
@Ecstatic_Hamster said in Peatbot:
I will post an invite link when it is ready for beta testing. I think you will absolutely love it.
OK, I would be glad to try it.
-
@Kvirion thank you.
-
@Ecstatic_Hamster Keen to try. You will post a link here? ETA?
Thanks in advance for working on this. When using popular LLMs, I debate how much to divulge and when/how to cue the LLMs about which particulars to remember and how to store/organize previous chats. The surveillance/privacy risks have led me to hesitate or write round-about prompts. Like many of us I would guess, I dislike the heavy-handed orthodox-Rockefeller-medicine-and Medical-Council-influenced boundaries/gate-keeping reflected in the way the popular LLMs are trained.
On health topics, I often repeat cues in my prompts over and over, like "include both orthodox and alt-health/nootropic-enthusiast perspectives" on topic xyz.
I wonder if you might include default tick-boxes (showing what orthodox medical-council-controlled journals say, what mainstream biology/physiology/biochem journals say, what alt-health/nootropic enthusiast communities have to say , etc., by category) while defaulting to returns based on intensive training on the Ray Peat corpus, the radio interviews, the email repository, RPF/low-toxin forum, peatatarian, bioenergetic forum. Sometimes i want to check only his articles or only his interview transcripts or only the email-response repository.
My search-engine/LLM wish list would include being able to easily select which sources should be included for specific searches, and maybe some user-controlled weighting or contrasting returns: e.g. 1 reply based on a mostly RP-corpus-weighted voice, 2 reply based on wider RP-influenced corpus, 3 contrast with mainstream/orthodox understandings of mechanisms of action, what's accepted, what's being questioned in emerging research, with some indication of types of journals/sources. I'm not sure the best categorization system. Just some ideas that have crossed my mind. Regardless, I look forward to seeing your design ideas.
