20 sur 20 personnes ont trouvé le commentaire suivant utile
le 14 mai 2006
Ray Kurzweil annonce dans ce livre très stimulant l'émergence d'un état transhumain, alliance de l'homme et de l'intelligence artificielle, qu'il peine à décrire, mais dont il pressent l'avènement proche. Il démontre de manière trés convaincante que la croissance exponentielle des capacités de calcul, au rythme jamais pris en défaut de la loi de Moore, et plus généralement des technologies GNR (Genetics, Nanotechnoly, Robotics), entraîne l'humanité beaucoup plus vite qu'elle ne l'imagine, toute habituée qu'elle est à raisonner de manière linéaire.
Conscient des nombreuses menaces associées à ce rythme inédit de développement, il considère les apôtres de moratoires ou autres principes de précaution comme irréalistes, voire dangereux et fait le pari que la technologie saura engendrer des parades.
Mais à supposer qu'il ait raison sur ce point et que le développement technologique ne sera pas entravé, ne fait-il pas preuve d'optimisme en pensant que la performance d'une civilisation ne dépend que de sa puissance de calcul ou de la densité de son réseau internet ? En n'abordant pas les questions de gouvernance, il semble faire confiance aux capacités d'autoorganisation du système qui sont loin d'avoir été démontrées ...
2 sur 2 personnes ont trouvé le commentaire suivant utile
le 26 août 2011
Ray Kurzweil's Law of Accelerating Returns, the exponentially accelerating rhythm of technological progress is obvious but to quantify the phenomenon is an obsession for him. This brings to a mistake when he says that the doubling of the capability of information technology occurs every year and that it means a multiplication by 1,000 in ten years and by 1,000,000 in twenty years. In fact 11 years will reach 1,024 and 11 more years will reach 1,048,576.
But Kurzweil starts with a mathematical example to explain "the singularity", a mathematic concept adopted by physics that he transfers into philosophy. His example is the function f(x) = 1/x. It is a hyperbole centered on the orthogonal axes of a Cartesian plane. For x = 0 the function is undefined, hence for x = 0 + n, n being as small as conceivable, and a dimension can always be cut in two, f(x) moves towards the infinite. In the same way if x moves towards the infinite f(x) will move towards 0: asymptotic growth. He doesn't consider when x becomes negative though then the hyperbole is perfectly defined, symmetrical to the first graph.
It is the old Achilles and the Tortoise Paradox, thus defined by Aristotle: "In a race, the quickest runner can never overtake the slowest, since the pursuer must first reach the point whence the pursued started, so that the slower must always hold a lead." (Aristotle, Physics VI:9, 239b15). In real life, the remaining distance between Achilles and the tortoise either is shorter than Achilles' arm, then Achilles will pick up the Tortoise, or it is smaller than Achilles' step then Achilles will overtake the Tortoise.
In fact Kurzweil ignores (two meanings intended) René Thom's Catastrophe Theory. In such cases of asymptotic growth x reaches the point of a qualitative change, and that is exactly what happens when x is nearing zero. In nature, no matter how small a particle of matter is, there is a qualitative threshold from one state to the other, from a particle to pure energy for example. That's what nuclear fusion and fission produce.
This beginning then falsifies the whole reasoning. There will be a point when technology is so developed that humanity as a whole will step over a limit. When is it going to be reached? For Kurzweil as soon as computers are as intelligent as man, or rather more, and he dates it. In forty tears or so.
Then Kurzweil evacuates the problem of language in about one page and one reference to Chomsky (p. 190). His man, and of course machines have no articulated communicational language. The whole theory is based on the consideration that intelligence is nothing but a problem solving mental procedure. The question is not whether it is genetic or the result of the functioning of the brain. The real question is what is intelligence and where does it come from. The author's insistence on the fact that evolution (producing intelligence) only starts with biology (p. 387), that intelligence is only human, that the rest of the material world does not even contain any rational element, that it is human intelligence, when equaled and overtaken by the intelligence of machines, that will take possession of an essentially irrational cosmos, this insistence shows his man-centered and even machine-centered ideology.
He cannot understand that from the very first instant we may consider, and that is not the beginning of the cosmos or matter - there is something before the big bang - the matter we consider is rational and has an architectural dynamic pattern. From this starting point, which is not the beginning of the cosmos but an arbitrary point taken just after the "big bang", the rationality of each state will produce that of the next. The question is how more than if. From the very first matter particles evolution will produce life. The evolution of matter particles from simple to complex will eventually produce the elements that will make life possible and then, along with this geological evolution, we have the evolution of species where haphazard mutations are selected by natural selection: useful or not for the survival of the individual or species.
Here Kurzweil falls in an enormous trap: language. Human intelligence is based on conceptualizing power. That power can only develop if the mind is able to label these concepts derived from the real referential environment, which means a language entirely invented by the mind in a situation when the body, as a side effect of bipedal running, produces three phylogenic hierarchical articulations in a context of social cooperation and communication indispensable for the survival of individual and species. Knowledge is not given, like with Kurzweil but it is a construct invented through assimilation, which requires a knowledge acquisition threshold implying motivation, transferences and cognitive strategies. All these are absent from Kurzweil's approach. If a machine could produce a father transference hampering or dynamizing its learning, I might become a believing convert.
Last remark: his inspiration, probably unconsciously, is recuperating some old trans-cultural religious concepts. For example the six epochs of the universe (what happened before the Big Bang?) is a very kosher number. "The Singularity will ultimately infuse the universe with spirit." (p. 389) The Messiah of the Old Testament and the Second Coming of the New Testament promise the same illumination, after the end of the material world, with inspiration from the Spirit or the Holy Sprit of God himself. John said it, or Ezekiel did. Kurzweil speaks of "a common Buddhist ontology [that] considers subjective - conscious - experience as the ultimate reality" (p. 388). Total ignorance of Buddhism for which man's lot is the permanently changing cycle of birth-life-death-rebirth, i.e. the concept "dukkha", and the possibility through "nibbana" to move out of it and merge into the pure energy of the universe. This vision of Buddhism seems to be out of Ray Kurzweil's consciousness.
Dr Jacques COULARDEAU