Both Google and Facebook are in a tight competition to see who will conquer the limits of artificial intelligence and virtual reality, and so far, they are both offering some trippy results, at least for the short-term.
Empowering artificial intelligence by gifting it with the capacity of learning on its own sounds scary and worrying, especially when considering the consequences over the years; for now, however, it means that we can enjoy and be slightly disturbed by what it can do.
Google has created its own neural network for their computer, which is basically an artificial brain that has – in theory – the same capacities that a human brain does. Designers weren’t looking for a way to create robots that can clean your house or perform laborious work; instead, they went on and tested the concept of dreaming and what it means for an artificial intelligence.
Facebook didn’t lag behind either, also creating a neural network, and results were extraordinary for both tech giants. It might sound weird that computers could “dream” while “sleeping,” but that’s exactly what researchers tested.
In Facebook’s case, the neural network generated images that made sense in 40 percent of the tests. What happened was that both companies had provided the programs with specific information, such as animals, objects and buildings, which became the database for their training.
The end result was to create computer-generated art – but the actual results took the researchers by surprise, as it was like nothing they’ve seen before: impressive, complex and vivid images that surpassed the separate bits of information that was input initially.
Google took it to the next level by having the computer act as if it was hallucinating and the produced images looked like the neural network was high on LSD. The masterminds behind the project, Christopher Olah, Alexander Mordvintsev and Mike Tyka, were both surprised and proud of what they created.
They released a statement explaining how the program worked; the neural network was taught how to enhance recognizable features, so if a cloud vaguely resembled a cloud, it would enhance it until it actually depicted a bird that had apparently appeared from nowhere.
Facebook’s endeavors went to the opposite direction, by attempting to created realistic images of actual objects. In around 40 percent of the experiments, their model of neural network managed to fool a human into thinking that the image they are looking at is a real object whose picture was taken.
It might not sound like much, but researchers were pleased, seeing that the results showed a computer’s ability to think, design, and draw these images – all on its own, with no human input. Both companies are proving that the industry of virtual reality and artificial intelligence is advancing, whether we like it or not.
Image Source: Pichost