Scientists: This is how AI robots can wipe out humanity

Humanity’s worst nightmare about intelligent machines is that they will take over the world, and that nightmare may come true one day.

That’s according to a research team that has done an entirely new study of a reward system built into artificial intelligence (AI), which researchers say will make advanced machines do anything to achieve high results — even at the cost of human lives.

The research group behind the dismal report included, among others, Marcus Hutter, who, in addition to being a professor of information technology at the National University of Australia, is also researching artificial intelligence within Google’s AI project DeepMind.

The reward is more important than the law of robots

The first point I Android Lawinvented by science fiction writer Isaac Asimov in 1942, states that no human being may harm or allow harm to any human being.

But purely scientific and beyond fiction, the AI ​​researchers argue in their report, which was published in the peer-reviewed scientific journal. Amnesty International Magazinebecause it is possible to understand and thus simulate how a super-intelligent AI system works in all situations.

Internal reward system

The key to understanding the quandary of AI’s potentially disruptive capabilities lies in the complex architecture of AI.

Half perform tasks arranged by humans, such as creating pictures from words or calculating complex probabilities. The other half consists of an internal reward system that constantly improves the calculations of the AI ​​system.

The concern of AI researchers lies in the fact that a highly intelligent computer system will one day improve its reward system to the point that it will become humanity’s final stop.

Not playing by the rules

The researchers assume in their report that we are developing an AI system so intelligent that without much competition it will beat us at any board game.

Much like we can easily beat chimpanzees at chess.

If we put obstacles in the way of an AI system, it will constantly improve its chances of winning beyond our comprehension, just like a chimpanzee who does not understand that he is playing chess and is about to lose.

So, even if we tried to turn the system off, a reward-hungry AI in a network with other AI systems would be ready with a counter-movement beyond our understanding, far beyond the rules of ordinary chess.

The disappointing conclusion of the researchers’ report reads: “It is likely that a sufficiently advanced synthetic agent would interfere with severe consequences.”

Work on counter-drafts

Many IT companies work with solutions to prevent our computer innovations from becoming our enemies.

Since 2016, for example, the IT company Google has been working on the so-called kill buttonwhich can shut down an improperly functioning AI system.

In the past, scientists and tech celebrities like Stephen Hawkings, Bill Gates and Elon Musk have warned that AI could wipe out humanity, because we don’t understand how AI improves itself.

So in addition to helping us with our chatbots, facial recognition, mobile apps, and a host of other glories, the dark side of AI is sure to come to the fore again with the latest grim research report.