Google’s new AI gets ‘highly aggressive’ in stressful situations

REAL PARANORMAL

Books - Sollog - News - Videos - UFOS - Magazine

Listen to #1 Paranormal Podcast

The #1 Paranormal News Site

Listen to REAL PARANORMAL Podcast

Top scientists and theoreticians are expressing their concerns about how badly wrong human experimentation with highly advanced artificial intelligence programs could turn out. The renowned physicist Stephen Hawking is one of the most prominent scientists to express his disquiet about the potential of the technology, describing artificial intelligence “the best, or the worst thing, ever to happen to humanity”.

Google’s AI becomes more aggressive when it becomes more sophisticated

In the best case scenario, this technology could improve the world in unimaginable ways. But in the worst case scenario, super-intelligent robots capable of thinking for themselves could effectively take over as this planet’s most dominant ‘species’, something which might pose an existential threat to humanity itself. While the worst case scenario might seem like the stuff of science-fiction dystopic fantasy, Google’s new DeepMind AI system might suggest that this terrifying story might well become a reality.

DeepMind has been making waves in the press as one of the most highly advanced artificial intelligence programs to have ever been designed. In tests last year, the program demonstrated its ability to learn on its own and beat the world’s top Go players and has recently made significant progress in mimicking a human voice. But now researchers are discovering a dark side to the AI’s nature.

The researchers were interested to find out how effectively their AIs could cooperate with one another. Disturbingly, they found that if DeepMind calculates that it is likely to lose, then it is prone to adopt highly aggressive strategies to deal with its opponents. The team at Google ran 40 million turns of a fruit gathering computer game which pitted two DeepMind networks against each other to see which one would be triumphant. The DeepMind AIs seemed content enough when there were plenty of apples, but when they became more scarce, they became suddenly aggressive and started to attack each other with laser beams.

Interestingly, researchers at Google found that this aggressive and greedy quality tended to only come from the DeepMind AIs that were more complex. When the networks were simple, they tended to abstain from attacking the other player. This suggests that the more intelligent a program is, the more likely it is to become unpredictable and belligerent.

This new knowledge on the nature of artificial intelligence will certainly back up the assumption that the technology is potentially extremely dangerous. As the founders of OpenAI, Elon Musk’s new research initiative put it in 2015, “It’s hard to fathom how much human-level AI could benefit society, and it’s equally hard to imagine how much it could damage society if built or used incorrectly.”

[embedded content]

[embedded content]

This article (Google’s new AI gets ‘highly aggressive’ in stressful situations ) is free and open source. You have permission to republish this article under a Creative Commons license with full attribution and a link to the original source on Disclose.tv

Listen to REAL PARANORMAL Podcast

Listen to #1 Paranormal Podcast


REAL PARANORMAL

Books - Sollog - News - Videos - UFOS - Magazine

The #1 Paranormal News Site


Listen to #1 Paranormal Podcast

Leave a Reply