Pony Preservation Project - /mlp/con 2020

How to contribute to the Pony Preservation Project:

• Anyone with godlike patience - Collect, organize, label, and maintain pony data in datasets
• AI and data people - Use the data, help fix errors, find good papers
• Anyone artistically motivated - Make original content, ponifications, artisanal shitposts
• Anyone that can keep up with what’s going on - Organize content, make sure people know how to help, write tutorials
• Developers - Make everyone’s lives easier, make collaboration easier
• Anyone able to reach an audience - Spread the good stuff, and make people love ponies and AI

This project is driven by the love of ponies and AI, and the contributions many anons. We can use all the help we can get. The more people we have working on this, the better our wAIfus will be!

Master doc:

• https://docs.google.com/document/d/1xe1Clvdg6EFFDtIkkFwT-NPLRDPvkV4G675SUKjxVRU/edit?usp=sharing
• Contains information on how to create an audio dataset, how to train a new character model, and how to get the synthesizer to do what you want

Panel:

• Slides: https://docs.google.com/presentation/d/1rHNNibckxfCoxZULs7hmaXDKcW7LgcgLYqEG5_ig4Vo/edit?usp=sharing
• Video: https://www.youtube.com/watch?v=j_vuTIP-WzA
• Contains an overview of the Pony Preservation Project and information on contributing
• Presented on Jun 13, 2020

AI blogs:

• DeepMind’s blog: science & futuristic capabilities possible today
• Google’s AI blog: what will be accessible in a few years
• Facebook’s AI blog: practical suggestions for developers
• Microsoft’s AI research blog: practical research (usually)
• OpenAI’s blog: great new ways of thinking about AI
• Uber’s AI blog: getting AI to handle greater complexity

AI dev tools:

• PyTorch, Julia, NumPy: These are currently the best frameworks for creating neural networks. Neural networks are hard to use properly! Follow this >>33987023 wise anon's guide. (https://desuarchive.org/mlp/thread/33963949/#q33987023)
• scikit-learn, Apache Lucene, Apache MLlib: Tools for creating decision trees and information retrieval systems.
• C4.5, C5.0, Classification and Regression Trees (CART), Random Forest: Algorithms for creating decision trees.
• Collaborative filtering, Latent Dirichlet Allocation, TF-IDF: Algorithms for information retrieval. Read up on TF-IDF before implementing you own!
• SpaCy AllenNLP, HuggingFace, scikit-learn: Tools for natural language (text) processing.
• Kaldi, Parselmouth, librosa: Tools for working with audio files.
• OpenCV, scikit-learn: Tools for working with image files.
• PredictionIO: A tool for creating recommendation engines.

Pony data for ideas:

• Greentext dumps, fanfic dumps, thread dumps, image dumps, show BGM, fan music, YouTube playlists, comment histories, compilations, "Best of" voting results, blog posts
• Show, community, fanfics, videos, images, games, plushies