Idea Blog: Machine Learning Applications
This is the first part of my idea blog. I have a large archive of ideas and designs that I am releasing, and this is the first of them.
I believe using coercion to defend ideas as one would for property is morally wrong. I however believe it is unethical (fraud) to falsely claim or imply one is the origin of an idea. If you gain value from ideas in this blog, I request you donate a portion of that value in an amount you feel is fair.
Machine learning to remove laugh tracks from TV shows, train on audio clips before and after laugh track addition, add as option on multimedia entertainment websites like Netflix
Machine learning to translate speech into another language using the same persons voice and tone. Basically, lyrebird.ai plus translation. Train on audio clips of bilingual voice actors reading same lines in 2 languages. Use for google translate and to dub every TV show and movie for customers.
Machine learning to sync audio to lip movement. Train on video clips with audio offset versus accurate. Useful for lip syncing in music videos, useful for in video players for correcting misaligned audio.
Machine learning in competitive video games that allow the user to enter an ELO/MMR score and get a bot that plays at that skill level. Imagine a chess opponent that is always at your skill level. Imagine multiplayer games with full servers.
Machine learning for phones to be somewhat better at guessing which way the user would like the screen rotated. Train on sequences of accelerometer data as input, guessing 1 of 4 orientations as output, with negative feedback if the orientation is changed again in less than 5 seconds, indicating the user did not desire the orientation selected.
Machine learning to skip commercials on TV with a single button. Start with database of commercials and various uncut TV shows. Write software to simulate commercials by injecting commercials into the uncut TV shows, and have the network train on guessing the start and end timestamps of commercials.