Google AutoML Creates Machine Learning Code That Is Superior To Humans

Google AutoML Artificial Intelligence System has recently produced a series of machine learning code. It’s efficiency even higher than the researchers themselves.

Obviously, this is another “human superiority” of the blow, because the robot “students” has become a “self-replicator” master. AutoML was developed as a solution in the absence of top-level programming talent for artificial intelligence. The team presents a machine learning software that can create self-learning code that runs thousands of simulations to determine which aspects of the code can make improvements and continue the process until the goal is reached.

GoogleNet architecture design diagram

Image Credit: Google

This is a great show of “infinite monkey theory“, but Google did not let a monkey hit the keyboard to create Shakespeare, and to create a self-replication programming machine, and these machines within a few hours of performance. It is better to work for a few weeks or even months than a human programmer.

Google AutoML Creates Machine Learning Code-1

Although it sounds a bit scary, AutoML does well in the programming of machine learning systems, far better than the researchers who created it. In an image recognition task and achieves a record 82% accuracy rate.

Even in some complex artificial intelligence tasks, since the creation of the code than the human programmer superior. It can mark multiple points in the image, the accuracy rate of 42%; as a contrast, the human software to create only 39%.

Google AutoML Creates Machine Learning Code-2

Of course, it does not mean “Skynet” or creepy “digital ghost”, because we are not in the “self-perceived machine” on the edge of the singularity. But that we have the technical potential of artificial intelligence and added an accelerator.

Google announced AutoML five months ago, in view of its ability to build in a short period of time than the researchers themselves better machine learning AI system, the results of the next year is clearly more worth the wait.