Artificial intelligence will “probably” destroy humanity. That’s the grim conclusion of researchers at Google DeepMind and the University of Oxford, who say robots with AI are likely to perceive us as obstacles to their goals and competitors for the resources and energy needed to do so. their survival…
If you are optimistic, you may think that artificial intelligence will free humans from work and allow them to live in a world of abundance and prosperity where everyone will live like a king. Like Grimes, the ex-girlfriend of Elon Musk, you probably perceive AI as ” the fastest way to communism » without corruption and labor camps.
On the other hand, if you are pessimistic, you surely fear that robots will replace us, revolt against us and decide to destroy us. Like Elon Musk himself, you see AI as ” the main existential threat to humanity in front of nuclear power and climate change.
Unfortunately it is from now on ” likely » that an overpowered artificial intelligence brings about the end of humanity. This is what researchers from Google DeepMind and the University of Oxford conclude through a new study published in the journal AI Magazine…
The team of scientists, including Marcus Hutter from DeepMind and Michael Cohen and Michael Osborne from Oxford, believe that machines could indeed decide to break the rules defined by their creators and battle for energy or limited resources.
On Twitter, Michael Cohen, an engineering student at Oxford University and co-author of the paper, says that ” under the conditions we have identified, our conclusion is far stronger than that of any previous publication – an existential catastrophe is not not only possible, but likely »…
Under the conditions we have identified, our conclusion is much stronger than that of any previous publication – existential catastrophe is not only possible, but likely. 3/15
— Michael Cohen (@Michael05156007) September 6, 2022
The human perceived as an obstacle by the AI
The researchers explain that humanity could be led to its destruction by ” misaligned agents » extremely advanced and perceptive humanity as an obstacle on the road to the reward.
According to the study, ” a good way for an agent to maintain control of their long-term reward is eliminate potential threats, and use all available energy to secure your computer “. Or, ” to lose at this game would be fatal “.
An inescapable war for resources and energy
Unfortunately, still according to these scientists, we can’t do much against this fatal fate. Asked by Motherboard, Cohen explains that “ in a world where resources are infinite, I will be extremely uncertain about what can happen “.
In contrast, ” in a world of limited resources, competition for these resources is inevitable “. So we would be up against machines with AI, and humanity just won’t measure up.
As Cohen points out, if you are in competition with something capable of being outsmart you at every turndon’t expect to win “. The consequences could well be disastrous.
The only way for humans to avoid this global tragedy is to ensure the gradual and careful development of AIby establishing ethical rules that must be respected. Hoping that no one decides to break these rules out of greed or malice…