Google begins developing an "Undo” button - in case of a machine uprising.

According to developers and scientists, the best way to stop a machine uprising is to develop a button that stops all machines at once. Google's DeepMind department, along with FoH, published a study that discusses ways to take artificial intelligence under forced human control in case of over development of neural networks. One of the best ways is to hit the "end" button in time for all devices.

Developers believe, at the current rate of growth of Artificial Intelligence, it will, in a very short time, be able to greatly surpass the human race in development. This level of development is commonly referred to as the Singularity. Confirmation of this hypothesis comes from an experiment that was conducted in 2013. A self-taught algorithm for the game of Tetris was able to independently determine that when you press Pause - the game process stops. Although it seems to be a small problem globally, in the long run it can lead to a collapse. DeepMind and FoH experts are of the same opinion.

To stop the potential revolt of technology against humanity, developers are working on an algorithm that, in the case of real danger, will be able to terminate the neural networks. As a company that writes software code and creates IT projects, we must assure you it is extremely difficult to do so, due to the limitations and autonomy of neural networks. All of its data is not processed on local devices, but on server-side power.

The best way to stop the invasion of a mad intelligence is to run a script on the servers that will block it, and will stop its work. Moreover, according to the idea, every inhabitant of our planet should have the right to run this script. Some scientists compare it to the "big red button.” In addition to the "Skynet" scenario in real life, scientists are analyzing simpler situations of debugging robotic technology.

The ability to safely shut down the system can come in handy in abnormal situations where the robot is out of human control," according to the company's report. Let's hope that the "robocalypse" situation will not occur, but it is necessary to develop technologies and scripts for post-emergency control of robotic equipment in advance.