Will AIs ever have a conscience?

That season 4 of western world is imminent: Dolores, Charlotte and Caleb might surprise us. Will the robot revolution end? We’ll find out on the day of the release date, June 26th. In the meantime, why not ask the question about that awareness ? It surrounds the show every season: from the first episode, where each presenter seems to be following a precise scenario, to the final season, where people plant their minds in artificial bodies.

Season 4 will be released on OCS on June 26th. © HBO, YouTube

How do you define consciousness?

Intelligence, consciousness, thought or even the living: so many terms for which the definitions do not find consensus. This is one of the mysteries that have surrounded us for hundreds of years, even thousands of years: Where does our consciousness come from and how is it defined? Several theories are based on this, but the majority of them distinguish between conscience and conscience self-consciousness. One is more oriented towards awakening and thinking skills and the other towards being different from other beings around us: it is this self-awareness that gives free will, moral values ​​or consideration of the world around us. One can quote the very famous ones Cogito ergo sum “, Where I think that’s me the mathematician and philosopher Rene Descarteswhich ascribes a form of certainty of existence to thinking.

Experiments have shown that other animals besides humans are self-aware. This has been shown in particular by a study elephants are part of it: You not only have the “examination of the mirror and Smudge”, in which they end up wiping off a smudge painted on their head after looking at themselves in the mirror, but also at that of the carpet. An Asian elephant managed to become aware that its body was resting on a carpet blocking the passage of a stick attached to the same carpet.

Science often works with biomimicry, that is, by being inspired by living beings. The digitization of human consciousness is therefore a logical step in the creation of an artificial intelligence designed to simulate or exceed our capabilities. Futura-Sciences met Jean-Claude Heudin, head of the research laboratory at the IIM (Institute for Internet and Multimedia), to understand the difficulties of such an undertaking. © Futura

At the side of Brain, the areas of consciousness remain untraceable for the time being. However, researchers have managed to map the brains of several different states of consciousness : during the sleepthe coma or when a person is awake. And they found differences in the interactions between regions: in conscious people they are more dynamic and complex, while in unconscious people they become simpler and only take place between regions that are directly connected. . As for the realms of consciousness, they would be in the brainstem for the zone of awakening, then in the prefrontal cortex. Finding all of these areas has the ultimate goal of mapping the brain. With a real-world map of cognitive functions and their location, researchers could then implant the same structure in a artificial intelligence.

Finally, a theory assumes that a Quantum Origin of Consciousness : The theory of the orchestrated objective reduction of consciousness, often referred to as the Orch-Or theory, first proposed by a mathematician Roger Penrose and University of Arizona anesthetist Stuart Hammeroff attributes consciousness to quantum computations in the brain. Controversial because the brain’s hot, humid environment would be too disruptive for quantum effects to survive, it emanates from electrical oscillations in microtubules, tiny structures found within the microtubules neurons of the brain. Researchers are currently experimenting underground to find almost undetectable radiation caused by these quantum fluctuations.

The key in western worldthe “daydreams”

But in western world, we learn from season 1, the consciousness of the hosts was built thanks to “daydreams”: they happen and repeat the events experienced at night while plunged into an artificial sleep. Contrary to our own dreams, Daydreams are made of real memories and are not a distorted reality, they do not contain incongruent elements. The next day, fragments come back to them, just like us, and little by little they become aware of themselves. But daydreams are more than that: the robots of western world, or the hosts, as they are called on the show, start hearing a voice in their head. That of Arnold, one of the park’s co-founders, who committed suicide to prevent its opening to the public, was convinced his creations were conscious.

The underlying assumption is, first, that dreams are paramount in establishing consciousness, not just memories already implanted in hosts. In fact, these memories create an identity for the robot, making it more human, but not enough to call its existence into question. On the other hand, the series speaks of a pyramid of consciousness whose first level is memory, the second level is improvisation (creation arising from a pre-defined program), the third level is personal interest and the top level is never explained, but it is assumed that there we find the famous conscience. Instead, the figure of Arnold, co-founder of the park western world, is based on the bicameral system, or bicameral theory of mind, put forth by psychologist Julian Jaynes, the idea that an its beginnings the man mistaking his thoughts for divine voices. This bicameral would have broken down little by little, and man would then have developed a real conscience. This hypothesis was disproved but made a lot of noise in the 1980s.

AIs are always evolving, will they pass the Turing test?

But for a robot or artificial intelligence that follows a program, the question of consciousness is different: how to ensure that if it develops signs of consciousness, it is real and not just the result of a computer program? Here is the Turing test comes into play, founded in 1950 by the mathematician and cryptologist Alan Turing. We get to know it in season 1: the park western world was authorized to open after passing the test. It consists in evaluating the ability of an artificial intelligence to imitate a human perfectly, to the point where it is impossible to distinguish it from a real person. For now, no AI actually passed the testbut it would be Google’s own Gato that would come closest.

Another Google AI has recently caused controversy: an engineer claimed that the artificial intelligence laMDA has reached a similar level of consciousness as we do. He based this on dialogues he had with the AI: in one of which she claimed to be afraid of their deactivation, which she sees as the equivalent of death in humans. But Google claimed that all of this dialogue was just improvisation on familiar topics, and that there is no trace of conscience there and the engineer was fired. This isn’t the first time something like this has happened since Ethics in AI remains controversial.

Interested in what you just read?

Leave a Comment