Google's AI is already capable of evolving without human intervention

7 predictions for the evolution of enterprise AI in 2018 | VentureBeat

Artificial intelligence is seen as something intangible, but the truth is that it is increasingly present in our lives. Now, Google has found a way to make its artificial intelligence evolve without the presence of humans by making the law of the strongest prevail. The program is called AutoML-Zero, and it is fascinating.

Mobile phones have been the gateway to AI in our lives, thanks to both the camera and certain processes of the operating system. However, it is in many more objects that we use daily... And in others that we will soon use in a massive way, such as autonomous vehicles.

The big business AI is much more complex and serves more than just helping us take better pictures. They are responsible for analyzing billions of data in record time and, in fact, if we weren't programmed that artificial intelligence, we wouldn't be able to do many of the calculations that are now done in minutes... Instead of hundreds of years " by hand "with conventional computers.

Having a 'mean' robot as your boss might improve your performance ...

Google is one of the companies that invests the most money in the development of artificial intelligence and, as we read in Sciencemag, one of the teams that operate with one of its AIs has created an algorithm to make artificial intelligence evolve similarly to that of a human.

This sounds as fascinating as it is terrible movie-making, but the shots don't go that way. The objective is to create an AI that is capable of creating protocols and algorithms that humans have not taught them to do, "thinking" logically as we would and evolving "on the fly", in addition to making other AIs evolve.

No, this is not Skynet and the Terminator is not going to emerge
The first thing that comes to mind when we talk about an AI that learns and selects according to the law of the strongest is, inevitably, Skynet, the Terminator network that is "fiction", but is deeply rooted in popular culture.

However, no, the AI ​​(at least, for the moment), is not self-aware and what AutoML-Zero does is select and improve other AIs. The main AI has a new algorithm that orders it to search and improve other artificial intelligence, but it is the "only" instruction imposed by humans.

What Is Deep Learning and How Will It Change Healthcare?

The main AI knows that it must solve a problem and, from there, it makes other artificial intelligence intervene in the process. When one succeeds, it "selects" it, discarding AIs unable to solve the problem. From that point, create new algorithms without human intervention to further develop the answer to the previously given problem.

It is a very important step forward for the development of artificial intelligence since we think creatively to create these algorithms, but AI is much more practical and, above all, fast in this task, so it "rambles" less and can find solutions to problems in which humans can stagnate.

Of course, there is time ahead for this AI to be finished. At the moment it is a project, but tremendously exciting.

Post a Comment