Est-il possible réellement de créer des êtres
artificiels ? Je veux dire : des créatures artificielles qui soient
si proches des êtres vivants que l’on ne sache plus faire la différence ?
Il faut bien reconnaître que la démarche classique qui
consiste à partir des technologies les plus récentes pour tenter de concevoir
des êtres artificiels donne des résultats intéressants, souvent utiles, mais
encore très loin de ce que la nature est capable de produire. Cela a été le cas
pour la technologie horlogère avec les grands automates des Lumières. C’est
aujourd’hui encore le cas avec les robots basés sur les cyber-technologies. La
question est de savoir pourquoi. Nous faut-il de plus gros ordinateurs, plus
rapides ? Est-ce que quelque-chose nous aurait échappé ?
Dans un post précédent, j’ai évoqué la thèse de la vallée de
l’étrange du Japonais Masahiro Mori. La vision traditionnelle est que plus on
ajoute de détails à un robot androïde, plus on se rapproche de l’humain et plus
l’empathie devient importante. Masashiro Mori a montré, qu’au contraire, plus
un robot humanoïde est similaire à un être humain, plus ses imperfections nous
paraissent monstrueuses. La théorie prévoit cependant qu'au delà d'un certain
niveau de perfection dans l'imitation, les robots humanoïdes retrouveraient une
plus grande acceptation.
De mon point de vue, cette vallée de l’étrange, où chaque
infime détail peut rompre la cohérence de l’ensemble, correspond dans les faits
à une barrière de complexité. Cette barrière est due à notre niveau de compréhension
(encore faible) des phénomènes complexes et à la nature même de nos
technologies qui ne sont pas (encore) aptes à permettre l’élaboration de
systèmes ayant le même ordre de grandeur de complexité qu’un être vivant. Nous
sommes capables de « simuler » très habilement mais « réaliser »
une créature artificielle aussi complexe qu’un être vivant reste une tâche
encore hors de notre portée, n’en déplaise aux partisans d’une singularité
technologique imminente…
L’approche qui consiste à franchir la barrière de complexité
d’en bas vers le haut en « sautant » au-dessus de la vallée semble
donc très difficile. Néanmoins, il existe une autre voie : celle qui
consiste à partir d’en haut et redescendre. Autrement dit, il s’agit de partir
du vivant et de le modifier en utilisant nos technologies. Les avancées
récentes des biotechnologies et des nanotechnologies semblent aller dans ce
sens. Ce chemin ne se heurte plus à la barrière de complexité qu’elle enjambe
avec aisance. Cela ne veut pas dire qu’il faille s’y engouffrer sans
réfléchir : heureusement quelques gardes fous éthiques permettront, du
moins je l’espère, de sauter sans tomber dans le précipice de la vallée des
monstres. Alors demain tous cyborgs?
Is it possible to
actually create artificial beings? I mean: artificial creatures that are so
close to living beings that we do no longer see the difference?
We must recognize that the traditional approach which is to use our latest technology to try to create artificial beings gives interesting results, often useful, but still far from what nature have produced. This was the case for the clock mechanical technology which allowed the creation of the great android automata during the 17th Century. It's still the case with robots based on our cyber-technologies. The question is why. Do we need larger and faster computers? Have we missed something important?
In a previous post, I discussed the “uncanny valley” thesis from Masahiro Mori. The traditional view is that if more and more details are added to an android robot, then it will be closer to human and our empathy will increase. Masashiro Mori showed, however, that when human replicas look and act almost (but not perfectly) like actual human beings, it causes a response of revulsion among human observers. However, the theory predicts that beyond a certain threshold of perfection in the imitation, humanoid robots would find greater acceptance.
From my point of view, this uncanny valley, where every small detail may break the coherence of the all, corresponds to a complexity barrier. This barrier is due to our level of understanding (still weak) of complex phenomena and the nature of our technologies that are not (yet) capable of allowing the design of systems having the same order of magnitude of complexity that living beings. We are able to "simulate" very cleverly, but "realizing" an artificial creature as complex as a living being is a task still beyond our reach, no offense to the supporters of the impending technological singularity ...
Thus, the approach to overcome the barrier of complexity from bottom to top, “crossing” or “skipping” over the valley seems very difficult. However, there is another way: one that is from top to bottom. In other words, it starts from real living beings and uses our technologies to modify them. Recent advances in biotechnology and nanotechnology seem to go in this direction. This path runs over the barrier of complexity with ease. This does not mean we have to rush into it: fortunately few ethical guardrails will allow us, at least I hope, to jump without falling into the abyss of the valley of monsters. So tomorrow, all cyborgs?
We must recognize that the traditional approach which is to use our latest technology to try to create artificial beings gives interesting results, often useful, but still far from what nature have produced. This was the case for the clock mechanical technology which allowed the creation of the great android automata during the 17th Century. It's still the case with robots based on our cyber-technologies. The question is why. Do we need larger and faster computers? Have we missed something important?
In a previous post, I discussed the “uncanny valley” thesis from Masahiro Mori. The traditional view is that if more and more details are added to an android robot, then it will be closer to human and our empathy will increase. Masashiro Mori showed, however, that when human replicas look and act almost (but not perfectly) like actual human beings, it causes a response of revulsion among human observers. However, the theory predicts that beyond a certain threshold of perfection in the imitation, humanoid robots would find greater acceptance.
From my point of view, this uncanny valley, where every small detail may break the coherence of the all, corresponds to a complexity barrier. This barrier is due to our level of understanding (still weak) of complex phenomena and the nature of our technologies that are not (yet) capable of allowing the design of systems having the same order of magnitude of complexity that living beings. We are able to "simulate" very cleverly, but "realizing" an artificial creature as complex as a living being is a task still beyond our reach, no offense to the supporters of the impending technological singularity ...
Thus, the approach to overcome the barrier of complexity from bottom to top, “crossing” or “skipping” over the valley seems very difficult. However, there is another way: one that is from top to bottom. In other words, it starts from real living beings and uses our technologies to modify them. Recent advances in biotechnology and nanotechnology seem to go in this direction. This path runs over the barrier of complexity with ease. This does not mean we have to rush into it: fortunately few ethical guardrails will allow us, at least I hope, to jump without falling into the abyss of the valley of monsters. So tomorrow, all cyborgs?
No comments:
Post a Comment