Google to upgrade video search with visual suggestions


Google search is soon going to be a lot more interesting as the tech giant utilizes some of the most consumer based advances in machine learning and artificial intelligence, commonly known as ‘suggestion engines.’

News from the Verge shows that Google searches will soon include visual ‘swipe left or right’ galleries where the technology will suggest individual shows and programs for users to watch based on their search entries and other data.

The new search functionality will also lead to various streaming video venues.

“Once you’ve found a show to watch, Google will let you play the content from streaming services you’re already subscribed to,” writes Richard Nieva at CNet, detailing how the new search will work. “That could include Amazon Prime Video, Netflix, HBO Go or others.”

Experts have long used suggestion and recommendation engines as a prime example to illustrate how machine learning works. The training set and test set policies that drive neural networks and related tools are especially useful in pinpointing user preferences or predicting trends based on past activity.

Notably, machine learning engineers are going beyond this now, to create deep reinforcement learning or deep Q networks that will teach computers how to play games based on experience. A computer will be able to play through small clips of Pac-Man or Shark Attack or some other visual game and learn what’s good, and what’s bad, though the value learning problem still applies. (simple premise? Try to win the game.)

The type of game theory that will be built into Google search is more of a passive style, but will still be pretty impressive, as marveling users wonder exactly how the machine knew what they wanted to watch. It’s all about weighted inputs, activation functions, and hidden layers.

Look for this functionality to come to the consumer stage relatively soon, to make our world even more eerily uberveillant.