Artificial intelligence technology has made an enormous progress over the last few years. So much that it even got some people worried about its potentially destructive nature. However, there is one area where it is still not as evolved as scientists would like it to be. That is its capacity to think and anticipate various events and to act in consequence. It seems like now, that one last impediment might become a thing of the past.
Speculative glimpse of the future
Researchers at Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Laboratory have succeeded in developing an algorithm which can predict what will happen in the near future. How does that work? Based on a still photograph, it can create a short video which shows what might happen following that particular scene. This speculative glimpse, if done successfully, might be the missing piece in the struggle to make artificial intelligence function like a human being. Scientists can then apply it to self-driving cars and many other devices.
Carl Vondrick, a PhD student at MIT and the lead author of the study said that the video they obtained from the still image is showing what the computer thinks might happen in the immediate future. He thinks that
“In order to predict the future, you must first understand the present. “
How they created it
Scientists created this innovative system using a method called “deep-learning”. More precisely, a branch of it called “adversarial learning”. This consists of two neural networks in direct competition with each other. The first one generates the video while the other one tries to distinguish between what is real and what is not. However, there is one problem. The video generated by the artificial intelligence are very short to be of any use, for now, at least. They are only one second and a half. Researchers are hoping that in the near future, they will become increasingly longer and more accurate. For now, this is what we have, but it is a start. So, people should not expect something like movies entirely made by artificial intelligence. That is still in the sci-fi domain for now.
One very important area where this technology might be indispensable in the future is the self-driving cars one. With a perfected artificial intelligence system, these cars will be able to function and anticipate events almost like a human driver. Accidents will be drastically reduced and technology improved greatly. The cars will be smarter and better prepared to react in unexpected real-life situations, like a careless driver or a pedestrian.
Another very useful thing this technology might be able to do is to fill gaps or to shed light on video anomalies in security footages. It will prove very useful in solving difficult cases which lack the clear proof of a security video. All in all, it seems like scientists are very close to developing the perfect artificial intelligence which will make a great difference in numerous domains. It is still work to be done, but we are the closest we have ever been to creating it.
Image source: here