Archive details

Discord archive

Neon Genesis Phijkchu

Lifecycle
July 17, 2020 – Present
Snapshot
Through August 30, 2026
Preserved messages
June 19, 2020 – August 30, 2026
Members
712
Messages
1,098,506
Channels
43

Archive navigation

vetting

Jump to date
Pins
Tinkerer_Red vetting
Replying to Spooky Phijkchu #spooky.131
So what's the angle you're taking to mature the AI
I've chosen not to use an "auto" encoder, and instead use an assisted auto encoder. So I spent about 8 months crawling the web for a wide veriety of words. then spent another 6 months crawling every synonym and antonym for all of those words, then the syn and ant of all those... etc.
Tinkerer_Red vetting
Then used an adjusted 1d boids algorithm to have every word move closer to similar words. repeated this until no word moved from a regular position in the list for 10 iterations straight.
Tinkerer_Red vetting
normally an auto encoder uses 2-5 values per word. If I can compress this down to just one and the encoder works properly then I can cut down the training time and calculation time by Nth degrees
Garbage vetting
Train an AI off of videos only featuring rob talking, i;e some lords episodes and pokemon nuzlockes to make an AI voice that kinda sounds like rob KappaHD /s
Tinkerer_Red vetting
Afterwards I'll be training it on a few months of messages between 2 replika AIs to givwe it a quick kick start, then training it on real human data between me and my exes since that is likely the largest dataset i have between 2 people talking. afterwards I will have a trained agent which can be applied to many more specific use cases
Tinkerer_Red vetting
Replying to Garbage #garbageatgaming
Train an AI off of videos only featuring rob talking, i;e some lords episodes and pokemon nuzlockes to make an AI voice that kinda sounds like rob KappaHD /s
I actually do intend to work on a voice module for the AI, but it wouldnt be truly synthetic, more like AI assisted smoothing of the phonetic audio clips.
Tinkerer_Red vetting
basically just take the single phoname, find it 100 times. stretch them into place and paste em all over each other. it'll sound like shit but the AI will smooth it properly (hopefully)
Tinkerer_Red vetting
convert the audio into a spectrogram, then have the ai do it's own smoothing filter. to train it you just have another ai which knows humans from robots, when it guesses the robot is a human 50% of the time then you have a good enough voice
Tinkerer_Red vetting
it would be quite some time from now as I am still working on improving training times and quality from the algorithm. Right now it is very fast, although it could easily be much much faster. for instance the activation function could be a unique one for this sort of data. since it's highly specific to text and not general data
Gent vetting
Make different versions of rob based on each era, we need a racist rob, a post-divorce rob, mindcrack betrayal rob, you get the gist
Garbage vetting
But then one day the AI started talking on it's own accord. Using the puppet pals as a false face. It was alive and it wanted revenge on phjikchu.
Tinkerer_Red vetting
The first iteration of the AI will have rig support for it's own puppet. So it'll be like a little cortona, since it's trained off my exes spea=ech it'll probably just end up doing thing to spite me. which should be fun to watch. learning that I have a type lmao
Tinkerer_Red vetting
I drop projects a lot and having too much scope can kill tons of projects. these of course are all just ideas. programming an entire model rigging system is it's own project. let alone hooking it up to an AI.
Garbage vetting
I was trying to think what would be the most incomprehensible but functional dataset, ideally it would be early 2000s video games and anime forums so you could get all that really cringy early 2000s posting style but in AI form, but so many of those websites are gone, so fanfiction sites would be the only massive archives of text that could train something with a similar mindset.
Tinkerer_Red vetting
dude... looking into neural networks written with out libraries, will return some amazing results. like every time you do find one it's in the most fringe as language you cant even read. so you have to cross reference it with other languages you also dont understand untill you get something you think might be readable
Tinkerer_Red vetting
Replying to Garbage #garbageatgaming
I was trying to think what would be the most incomprehensible but functional dataset, ideally it would be early 2000s video games and anime forums so you could get all that really cringy early 2000s posting style but in AI form, but so many of those websites are gone, so fanfiction sites would be the only massive archives of text that could train something with a similar mindset.
gaia online message boards with role play threads from early 2000s

This archive is read-only — snapshot through August 30, 2026

Channels:
Density:

Jump to date

Local bookmark

Choose a color

This marker stays in this browser. Your choice becomes the default for the next bookmark.