AutoML even created a better self-learning code system than the one who made it. When identifying images, it reaches an accuracy of 82%. Even in some complicated AI tasks, self-generated code is more than human when it comes to marking pixels as accurate as 42%, while human-generated software is only 39% accurate.
This is not the first phase of Skynet or anything so scary, nor is it self-conscious machines. But we are making great strides to explore the potential of AI and promote technological progress.
Google just announced AutoML 5 months ago. This is a great milestone when AI can create machine learning system better than human, in such a shorter time. In addition, they are also willing to share their research as Google's AI researchers often do.
Perhaps you are not the only one who wants to see what Deep Mind can do when creating a Deep Mind itself better.
Progress in AI has only just begun and the future is very bright. AutoML is just the start of the new generation of leraning machines. Future machines will not only learn, but also update and create programs that can solve unforeseen problems.
Hopefully in the future, AI can take on time-consuming tasks like programming and let people do things that machines can't do.