Google is working on a « button » to stop AIs that would like to attack humans

Researchers at Google are currently working on a device to disconnect an artificial intelligence to prevent it from taking over humans.

Artificial intelligence frightens as much as it fascinates. With the algorithms implemented which develop more and more, Google forces itself to envisage a way of preventing the loss of control of the Man. In some areas, artificial intelligence is already surpassing humans. Last March, for example, Lee Sedol, ranked third in the world in the game of Go, was beaten by a Google algorithm. AlphaGo“.

Google has been interested in and developing artificial intelligence for a long time. The firm even invested just over $600 million in 2014 to acquire the services of the British company Deep Mind Technologies, which specializes in this cutting-edge sector. The main objective is to set up autonomous machines and create algorithms that can automate many processes. But Google is also aware of the need to be careful with the rapid and constant development of these programs. The company is therefore seeking to create at this very moment a mathematical system that allows a person to interrupt the operation of an artificial intelligence at any time in the event that it becomes malicious.

DeepMind is working closely with researchers at the University of Oxford on this device, which does not currently work on all algorithms, which is problematic and is currently hampering the development of certain AIs. Because an artificial intelligence is by nature capable of learning and evolving by itself. Researchers must therefore wait for certainty on the operation of this “red button” in a way, this emergency stop function.

The scenario of the film Ex Machina shows precisely how an AI can turn against its creator:

Sources: MIRI; 01net

Laisser un commentaire