The paper which was published last month in the AI Magazine, a peer reviewed publication, suggests that AI could likely destroy humans, arguing about a possible existential threat due to the technology.
Why scientists believe that AI might become an existential threat
The researchers point to the AI development networks called GANs or Generative Adversarial Networks. The paper claims that a ‘cheating strategy’ may be developed by an advanced AI in the future in order to achieve highest rewards even when it harms humans
The scientists make the claim also based on the fact that we live in a world with finite resources. If the world had “infinite sources, it would be uncertain how things pan out. However, with the finite resources available to us, “there’s unavoidable competition for these resources” and humans “shouldn’t expect to win” when in competition with “something capable of outfoxing you at every turn”, the paper’s co-author Michael K. Cohen of the University of Oxford was quoted as saying by Motherboard.
The scientist also argues that the “insatiable appetite for more energy” of such an advanced AI will “keep driving the probability closer and closer”. They argue that such an AI might ignore the essential tasks humans assign as per their need and do its own tasks in order to gain its rewards.
The paper says that humans will be stuck in a zero-sum game between survival and need for technology in such a scenario and “Losing this game would be fatal”. The researchers advise that such advanced AI should not be built unless there is a guaranteed method in which we can stay in control.