The dangers of artificial intelligence are a classic of science fiction, but if the one who launches the alarm is the famous physicist Stephen Hawking, maybe it's really time to consider it as a real problem.
"The development of a strong artificial intelligence could lead to the end of the human race"
he explained in a BBC program.
"Primitive forms of artificial intelligence (weak AI) have proved to be useful, but I fear for the consequences of creating something that can equal or surpass human beings. It can take off and reprogram itself at an ever-increasing speed and we risk to be left behind: "Humans, that are limited by a slow organic evolution an that can no longer compete with intelligent machines, might be overwhelmed"
In the Social Networks there has been lot of discussions: the first was Stephen Hawking, then Elon Musk and more recently Bill Gates. All these brilliant people have suggested that artificial intelligence (IA) is something to look carefully at, to avoid developing it into an existential threat.
But what type of AI are they speaking about?
It is difficult to find any details about this threat, but thinking about what Dr. Hawking said (I have quoted only a few words of his discussion)I've reached some conclusions.
It is tempting to suggest that a protection would be the introduction of empathy in artificial intelligence. But I think it might fail for two reasons. First, most of the humans have empathic skills, though they continue to be unpleasant, uncivilized and brutal towards their counterparts and every other living organism on the planet. Secondly, it is not obvious that true/strong artificial intelligence is something that we can develop "step by step", maybe we should allow it to evolve on its own.
We can already create an AI that can learn by himself (Automatic learning), an example is Galileo, a computer built in USA that studies physics observing some phenomenons and thinking about how they works.
Perhaps we can create the general conditions from which intelligence can emerge. At first glance it may seem grotesque, like throwing a bunch of spare parts into a box and hoping that a new bicycle appears. It is certainly not a scientific way to treat artificial intelligenc. But if intelligence is an evolutionary property that emerges from very, very complex system, could it happen? Maybe.
It is a bit stupid to sit and think about creating the ideal circumstances for the appearance of a new kind of intelligence, since we don't know what type of circumstances they are.
But maybe we are really creating those conditions. Automatic learning can only be a piece of the puzzle of artificial intelligence behavior, but what if this intelligence will live through the Internet extensions? Data treasures, apps, algorithms that control packets of data, sensors (from GPS to thermostats to traffic monitoring ) and an immense numbers of "pieces" that "speak" to each others directly or indirectly.
Will something like a strong artificial intelligence born from all this things? I do not know, and nobody knows. But this is a situation that has never existed in four billion years on this planet. And this brings us back to the question of a possible threat from the artificial intelligence.
If this is the way to create a strong artificial intelligence, the immediate danger is simply the fact that a vast array of human beings now rely on the Internet ecosystem. Not just for the way we communicate or we find informations, but also how our food supplies are organized, as pharmacists order drugs, how our planes, trains, trucks and cargo ships are scheduled, how our financial systems work.
An emerging strong artificial intelligence could bring the chaos like a child who disrupts your socks drawer.