As I've been building out my machine learning pipeline, I been feeling like that episode of Star Trek Voyager where the EMH Doctor steals the mobile emitter from the 29th century, and later accidentally loses it to residue from Borg Nanotechnology who promptly extracts the technology there to bootstrap itself to something that becomes more powerful than anything seen before. Or like that episode of Mr Manhattan rising for the first time: https://www.youtube.com/watch?v=sNXohNU3tWo
I'm rapidly approaching the boundary of what is possible for computer science and machine learning, and redirecting that horsepower on itself in ways that were done few times before, if ever. I suppose this is the holy grail, like fusion researchers need to get over parity so that the reaction goes exponential, like how the first piston engines were able to capture the detonation of fuel in a controlled way, in the same way we need to contain, capture and output machine learning in a controlled way.
Automating all forms of human labor to robotics using computer vision and machine learning may be the key to unlocking Star Trek levels of prosperity, bringing unforseen consequences as did agriculture 10000 years ago, the industrial age 150 years ago and the computer age 20 years ago. With 9 of every 10 people not really needing to work because expensive computer equipment and robotics can do the job better and more cheaply, then the economics of Star Trek or outer limits Utopian ideologies start to be possible, as is explored when replicators become widespread, the dominant currency of life transitions from a measure on human labor to a measure on quantity of raw materials collected and net yield on energy collected from the sun. When technology enabled one man to meet the needs of a thousand others in agriculture, food went from being the only thing to practically zero cost. This can be done for all professions.
Tesla Developer lead Andrej Karpathy's CS231n Winter 2016 Lecture on Convolutional Neural Networks is the best I've seen: https://www.youtube.com/watch?v=LxfUGhug-iQ&list=PLkt2uSq6rBVctENoVBg1TpCC7OQi31AlC&index=7
I want to put my recent machine learning degree toward the goal of advancing artificial intelligence, computer vision, and building interface software. High hanging fruit would be between computer disk storage and the human short-term long term memory via potassium ion wave form voltage gated ion channels. Lowest hanging fruit is probably using the latest in machine learning software on beefy TitanX GPU's to make inefficient physical object markets and point of sales more efficient through data analysis, arbitrage, robotics, computer vision and data interpretation/discovery.