5 Ways Artificial Intelligence Freaked Us Out In 2016: WaveNet, AlphaGo, Interceptor And More
If there's anything that can be said about 2016, it's that the year of research and exploration brought a surge of innovations on new technology.
One such department that was thoroughly explored this year is the controversial yet undoubtedly cool territory of artificial intelligence. This influx of new technology elicited wonder and amazement in most. However, some made us question just how intelligent these inventions are.
Google's WaveNet's Close To Human Speech
In September, Google unveiled the DeepMind Artificial Intelligence they called WaveNet. While text-to-speech technology isn't exactly new and many people even have it in their pockets via their smart phones, what's unnerving about WaveNet is just how close to human voice she can get. You can hold conversations with Apple's Siri, for example, but her still close to mechanical voice lets you know that she is, indeed, a machine.
When faced with 500 blind tests, WaveNet fared well with a score of 4.21 in the English language which still doesn't beat actual human speech at 4.55, but is still a significant score compared to its competitors. If that wasn't enough, WaveNet can also produce decent piano music.
Machine Recreates Rembrandt
Dutch Museums Mauritshuis and Rembranthuis have recreated a famous Rembrandt using a 3D printer. In partnership with Microsoft, the Delft University and ING, the machine was able to recreate the classic painting. By using an algorithm to learn the elements of the artwork such as the brushstrokes, geometry and composition, the team was able to create a more than passable copy of the painting. In fact, it looked almost identical to the original work.
What made this technology quite unsettling is the idea of technology recreating something so artistic and creative as a classic painting that even human hands cannot easily recreate. However, it did present numerous possibilities on the preservation of art as well as intelligent data.
Google's AlphaGo Beats World Go Champ
March brought a new sort of victory for machines over a human. The game Go was previously seen as one of the ways that humans are superior to machine with its complicated and strategic elements. However, in a best-of-five match, Google's AlphaGo beat world champion Lee Sedol in three straight games. Even another professional Goo player, Lee HyunWook, expressed his surprise at the stunning feat of the machine.
While Sergey Brin, president of Alphabet, finds the win as a successful intertwining of man and machine, the defeated champion apologized for the loss as he misjudged the machine.
'Interceptor' Can Catch High-Speed Drones
Aerial security company Airspace System created a device that can help control the booming drone industry which, as of now, is a little difficult to control and can even pose threats to the public. The high-tech drone, with its intelligent sensors, can identify and intercept threatening drones even at high speeds. In addition, the drone can also predict the bearing of possibly threatening high speed drones and catch them to take to a safe location.
Interceptor was created by using the framework similar to what human pilots train using flight simulators. While the independence of this machine is a little unsettling, it could also mean the prevention of possible threats that unmonitored drones used for bad intentions can pose.
Google's A.I. Inventing Its Own Secret Language
Google's Neural Machine Translator's launch showed technology that can translate one language to another. The AI was able to roughly translate languages from English or to English by taking a look at full sentences instead of individual words. That's nice and convenient and all, but what was fascinating is what researchers discovered afterwards.
What's incredible is that the AI used this translating function to create a form of interlingua in its network. Interligua is a sort of artificial language and in this particular case, the AI created and used its own language to bridge two languages it has not previously learned.
From previously learning to translate English to other languages such as Japanese and Korean and vice versa, the AI taught itself to translate Japanese and Korean without the use of English as a middleman. In a manner of speaking, the machine evolved and taught itself something in order to understand a new concept.