Google and Facebook have been working on testing the limits of virtual reality and artificial intelligence. While some have worried about what that could mean for computers and their powers in the long-term, in the short-term it’s meant some pretty interesting pieces of technology and research emerging from the space. That’s why Google went ahead and created a neural network for their computer, or in simple terms, a brain that functions the same way – in theory – as a human brain does.
The goal wasn’t to actually get the computer upright and cleaning your house, or going to work for you, but rather, to actually see if a computer system built like that could dream in the way that humans dream. The idea of a computer seeing and recognizing images, while it’s asleep, might seem strange to some, but on paper it’s the exact same concept as what humans experience when they sleep at night.
Most impressively, the results were really astounding for both Facebook and Google. Facebook found that 40% of the time, their computer would generate an image that was recognizable to humans. In both instances, the companies put together and trained these computers with specific pieces of information, objects, and animals in mind. The result was aimed at creating art that would be amongst the most interesting that any human has ever seen, which was put together by a computer.
The images were absolutely impressive. They were complex, vivid, large, and created things that simply hadn’t been imagined to that point. Google definitely took this to another level though, with their intense research and results. They went ahead and actually made the computer hallucinate, and in doing so, they created some seriously impressive images.
Alexander Mordvintsev, Christopher Olah, and Mike Tyka, all of Google, were the individuals behind the project, and were incredibly proud of the results. The team said in a statement regarding the research that, “If a cloud looks a little bit like a bird, the network will make it look more like a bird. This in turn will make the network recognize the bird even more strongly on the next pass and so forth, until a highly detailed bird appears, seemingly out of nowhere.”
On the other hand, Facebook pointed out that, “Around 40 percent of the samples generated by our class conditional LAPGAN model are realistic enough to fool a human into thinking they are real images.” While that might not be an overwhelmingly solid number, it’s pretty significant given the fact that this research relied on a computer to think, create, and draw these things, which otherwise would take human interaction to create.
This simply shows the direction, at which the industry is moving. There is a pretty significant amount of research available, and as we continue to expand on what computers can do alone, we’ll clearly see an increase in AI usage. This will either be a great thing, or a terrible thing – depending on your perspective, and preference with artificial intelligence.