Diary of a Robot Breeder – Episode 3
Artificial intelligence is not our enemy. It sends back to us the face of my contradictions.
This is a remark that comes up often in my conversations about our media robots: “Alright, ok, you made them cute your robots. But, in the end, wouldn’t you be helping to create Skynet?“
For those who never watch sci-fi movies (you should, actually), Skynet is a concept from the movie “Terminator” (the one where the robot soldier Arnold Schwartzenegger knocks a pump gun in his hand on every door, saying “Sarah Connor?”) (because he is looking for Sarah Connor who is kind of the last hope of humans). Skynet is a kind of omniscient artificial intelligence which, in the future (well, in Terminator…), had the rather logical idea of reducing humanity to slavery.
That was in the 80s. Back then, it made us laugh, Skynet. Except that since a robot beat a human in the game of Go, then in poker, we are talking more and more about intelligent robots capable of surpassing, even of deceiving humans.
Suddenly, we are talking more and more about Skynet.
So when someone points out to me, I avoid defending my robots and reassuring other members of my species about artificial intelligence. Because it is also one of the objects of the Flint project: to question our relationship to robots.
Well, afterwards, don’t panic, Flint’s robots are baby robots. They are relatively simple, compared to their big brothers Deepmind (Google) and Watson (IBM). They are obviously not racist like they were Tay (Microsoft). They can’t (yet) converse like Alexa (Amazon) and Home (Google).
However, they have one thing in common with most other artificial intelligences: they enjoy a certain autonomy. This is the price of intelligence: freedom.
So that you understand correctly: the mission of the robots of Flint is to determine, among the thousands of articles on the Internet, which are of quality and could interest their readers. We know what methods they use, but we don’t know why they are making this or that choice. It is one of the side effects of neural networks, one of the technologies used by artificial intelligencee which tries to reproduce the functioning of the brain.
That’s why, in the Flin projectt, each robot is supervised by a human. Not to impose selection criteria on it, but to educate it and teach it to be autonomous, without letting it drift.
In some situations, especially the most complex, the AI chooses the aggressive option and destroys its opponents.
This autonomy is essential for intelligence. But it is precisely she who is scary.
Because if the robot becomes more and more autonomous, and more and more intelligent, will it one day realize that, ultimately, killing humans is more effective than letting them live?
A recent study conducted by DeepMind engineers from a video game shows several scenarios borrowed by artificial intelligence when it is confronted with the management of resources (in this particular case, apples…). Basically, his behavior is quite close to that of humans, which is quite hopeless. When resources are scarce, the AI can choose to collaborate or eliminate the others … In certain situations, especially the most complex, the AI chooses the aggressive option and destroys its adversaries.
Demonstration in this video, which is quite incomprehensible, but rather cool to watch:
You will notice that we are still at the stage of apples in the shape of a green square in the brains of robots. But it’s a start.
Already, 30% of Americans say they’d rather be led by an AI than a human boss. It reminds me of Voltaire’s dream of enlightened despotism : an objective leader, not corrupt, who would probably make the most effective decisions. Like a robot. Yes, but according to what objectives? What if in order to achieve his ends he becomes aggressive? Should we provide a “stop” button in case the robot gets out of our control?
My robot is not my enemy. He sends me back the face of my contradictions.
For now, robots are not dangerous. Nor psychotics. One of the great specialists in the subject, Laurent Alexander, explains it very well in an intervention made in the Senate on January 19 : it distinguishes “strong” artificial intelligence from “weak” artificial intelligence, ie without self-awareness. Which is the case with all artificial intelligences today. They perform very specific tasks, often impossible for humans, but they are not able to “think” in the sense in which we understand it. In the meantime, therefore, they are very useful to us.
The most immediate impact of robotization on the human species will not be Skynet, but first the massive destruction of middle or working class trades, In the coming years. In absolute terms, the fact that robots are replacing us in difficult, automated, mind-numbing, and meaningless jobs is a good thing. Ultimately, ideally, they will give us time to think about ourselves, our future, so as not to act in haste.
New professions will arise for humans. New resources too.
But are we ready? Is the education and training system ready? And what about politics? Why is this so essential subject absent from the presidential debates?
In fact, the robot is not the enemy. The enemy is us.
The robot reflects back to us the face of our contradictions.
The robot is like the demons of legends, the “daimon” of Socrates, the “familiar” of homes, “The entity, animal or spirit, sometimes imaginary and invisible, to which men turn to seek advice or obtain services”… Goethe’s Faust. It is useful, it can even save us from self-destruction in an increasingly complex world. But it can also be dangerous if we are not careful. To be afraid of it is to be afraid of us. Fear of our choices. Or rather our non-choices.
So is it necessary, as Elon Musk advocates, merge with artificial intelligence so as not to be overwhelmed by it ?
Or should we first of all reposition ourselves as humans? To regain control of our future is first of all to ask ourselves this simple question: what is it to be human?
Benoit Raphael is an expert in digital and media innovation, blogger and entrepreneur.
He is at the origin of many successful media on the Internet: Le Post.fr (Le Monde group), Le Plus de l’Obs, Le Lab d’Europe 1.
Benoît is also co-founder of Trendsboard and robot media Flint.
FrenchWeb mobilizes and organizes the seventh digital employment week (# SEN7) throughout the week (portraits, interviews, job offers, round tables, etc.). The #SENFW operation, intended to promote employment opportunities in this sector, takes place every 6 months.