An artificial intelligence creates an enhanced software that puts everything created by humans in the shade.
To many people, this still sounds like music from the future: machines that are capable of independently creating other machines and computer programs, thereby continuously improving themselves. What for many is a horror vision from the Terminator movies or the Omnic invasion in Overwatch, has now taken its first step into reality.
Raymond Kurzweil, who has been working as Director of Engineering at Google for some time, has now achieved a small sensation with his team.
The developers at Google set their AI “AutoML” an exciting task. It should closely examine the components of AI NASNet and improve them with more effective versions. NASNet is basically a recognition software that is supposed to identify specific objects from videos in real time, such as people or cars.
The AI AutoML eagerly took on the tasks and made countless small changes to NASNet, making the system increasingly powerful.
The end result is that NASNet now recognizes objects 1.2% more successfully than all human-developed recognition systems and operates 4% more effectively. This marks the first time that an AI has managed to improve another AI more than humans have been able to so far.
This development is indeed a huge breakthrough in the development of AI, but it also raises many concerns that have previously been dismissed as strange fantasies of sci-fi nerds. After all, what machines can cause when they take control and continuously improve themselves has been imagined by us for decades through movies, series, and books.
Skeptics think: It will probably only be a matter of time before Skynet nests itself in all our computers. Whether that will be as entertaining as a movie remains to be seen.
What do you think of this development? A great thing that finally brings the next big step in technology? Or is this the beginning of the end of humanity?
