What was Tay?
Rip archives of 4chan.org, 2chan.org, 8chan, 420chan use as the teaching material for chat? (Somewhat done)
Gather listings of conversations, and responses. For given input A try to predict output B. (4billion conversations from Cleverbot, done)
Use IBM Watson for sentence understanding (NLP)?
Useful, but not really…
Rip videos of LillyPichu? for audio feed into the voice print for text to speech.
Gather voice samples split into collections of vocal sounds. Take groups of three of these sounds and feed into neural net. With output being one noise, Match noise to snippet from actual audio generating hamming weight of audio as back-propagation?
Neural Network Design (RNN):
Input nodes: Audio Snip
Middle node: Text of audio
Output Node: Audio clip of text
LSTM? Or 3 sets.
Gather faces of common Japanese and Korean pop idols, preform a merge of there facial features. Use that on a animated model to create video stream. (Obtained Deep fake Scripts have working POC)
Stream content to twitch as “Just Chatting”
Is it possible to represent language in multiple dimensions? But perceive them in lesser ones? Like stars in a night sky, each word having a brightness a position in a 3D space. But relatively to us these points shine differently to everyone. Like your perspective would be that of looking up at the sky. Then our sentences, form constellations in this 3d space. Now I wonder what would happen, if we took conversations as 3d polygons, and tried to predict the transformation of the object.