Looks impressive. Identify "things", label them as "Vectors", associate Vectors to people based on likes and dislikes, cross languages. Deep learning using Neural Nets. Training each layer at a time. Some history of AI, "Conspiracy of AI". AI momentum caused by paper by professor at MIT, stated lack of progress. Neural Nets were popular for a while, then winter, then resurgence. Lots of work done in Canada.
Bigger data sets. Cheaper costs (hardware, software).
Convolutional Neural Nets