domingo, 12 de outubro de 2014
RAY KURZWEIL AND "THE SINGULARITY"
Forgive me for the long text, but, as I think nobody will read this, will not make any difference ...
For someone who studies physics and astronomy, it is very easy to get lost in daydreams, thoughts and ramblings about the future. How will the world in 50 or 100 years?
Well, I'll try to do a simple analysis of what I have read on the subject.
A computer will write the best novel of all time in a second, solve the greatest mystery of science in a tenth of a second and discover the meaning of life, the universe and everything in less time than it takes you to finish this paragraph . Many of us will probably be alive when that day comes. The problem is how to be the next day ... This is what we see now. Welcome to the real Matrix.
It's been 15 years that Deep Blue, a supercomputer from IBM, beat Garry Kasparov, the trump card of the chess team's humanity. At the time, some people that little credit to the intelligence of that machine because the game is highly mathematical - science for which computers have more than natural ability.
In practice, there would be a great "intelligence" there. Only a big calculator. The computer could then even be geniuses, but they would never know "think like a human." Today that vision no longer makes sense. Watson, another IBM supercomputer, managed to win in 2011 the two best human Jeopardy players in a game show on American television. It is a game of questions and answers that requires participants a hell skill with language. Are questions such as "This object, even when it breaks, is right twice a day. Which is the object?". "A watch," replied Watson, with the same rapidity with which a calculator gives a result. And so with dozens of other questions along this line. "That did not happen, mind you, because some scientist put that information in the right way," says American Ray Kurzweil, one of the most celebrated experts in artificial intelligence. Indeed. Watson did all that before attending the game was read thousands of books and encyclopedias - including Wikipedia, though he was not connected to the internet in time to answer the questions. Everything had to be "head" even. Brave. And enough to say that Watson is to Deep Blue as a puppy is to an ant. It is an infinitely superior intelligence - but not arrive at the foot of human.
But some experts to our advantage that should not last longer so long. This is the case of Kurzweil. For him, the chips should overtake us until mid-century. That's when toparemos with what he and other scientists call a "technological singularity". Let's see what this is.
In physics, "uniqueness" is what happens inside a black hole. A black hole forms when a star implodes and begins to be destroyed by their own gravity. There comes a time when it leaves the star. Only gravity. In fact, a speck in space where gravity tends to infinity. This dot is what physicists call "uniqueness". Over time, the singularity will suck all you have around you. The mass and energy of the objects she swallows leaves the "strongest" black hole. The power of suction increases it, he swallows more stuff, get stronger ... In short: the growth of the black hole happens in an endless spiral. With the "technological singularity" would be the same thing. A more intelligent machine that humanity itself would create even more sophisticated machines without human programmers. Artificial intelligence grow on their own, like the suction power of black holes. This process would tend to infinity, with machines giving birth incredible machines that then create even more fantastic machines. And we would be just watching.
This is only a theoretical possibility. But perhaps we are now giving even the first steps to achieve this reality. And for a simple reason: build machines increasingly the image and likeness of our brain. Human intelligence has been produced by natural selection. Throughout evolution, our head grew to deal with challenges that our bodies would be unable to win - try to run from a lion on the African savannah for a more complete reference of what we speak.
The survival of human beings depended primarily on developing the ability to create extensions of our bodies - tools - while having the chance to especializarmos in the use of them. We can therefore summarize the human brain into two basic qualities: processing capacity (imagine the right tool) and plasticity (to adapt to the use of such a tool).
First let's talk of processing capacity. What is the computational performance of our brains? IBM researchers have made this account and estimate that the brain is capable of achieving 36.8 petaflops - or 36.8 quadrillion operations per second. This equates to roughly 1 million PCs working together.
Coordinate the work of many chips is still impossible with today's technology. But supercomputers have advanced a bit. In 2012, the Sequoia, the IBM has achieved 16.32 petaflops - almost half of the human capacity. Still, it does nothing as amazing as we do - things like thinking that we are alive. Why? To answer this question, enter the second basic quality of his head: the plasticity.
Your brain is highly malleable, adaptable. As the nervous system continues to develop, various brain regions will assume different responsibilities. It is known, for example, that there are certain areas for language processing. There are also specific parts of the brain that control different parts of the body such as the feet and hands.
But best of all is that it gives the brain to adapt as needed. It plasticity which allows, for example, a subject becomes large pianist. Brain areas related to hand expand greatly on who trains piano for many years. "Our brain is a system designed to learn, to shape the interaction with the environment," says psychologist Steven Pinker of Harvard, a leading expert in the workings of the gray matter. The best electronic brains work basically like today. Ie: they can learn. "The techniques that have evolved in the field of artificial intelligence techniques that are similar to the brain uses," says Kurzweil. "And that did not happen because this field of research was copying the brain." Ie: it was a happy coincidence. A coincidence that allowed machines like Watson, able to learn on their own when they read a book - like you.
As the plasticity of electronic brains increase, it is possible that an hour they become as complex as the brain you carry. Kurzweil, who is the greatest harbinger of this thesis, estimates that the machines will reach an equivalent to human intelligence in 2029 Yes, exactly 2029 He arrived at this number based on mathematical projections on the evolution of processing capacity. Of course the forecast is controversial - for many researchers, crimp the year that something so unpredictable and unusual should arise is madness. Point. Anyway, the accounts of Kurzweil, the singularity itself will begin in 2045, when a single computer is smarter than the whole human race.
A machine with such power would be as fascinating as dangerous. For one, she would be able to coordinate and execute all activities assigned to us today, how to write a great novel or unify into a single theory to quantum physics (which governs the subatomic universe) and relativity (which dictates the order in the world of great things) - something that Einstein died trying to do. All this in a matter of seconds. Beautiful. On the other hand, what is the level of respect that such a machine would have for us? We're talking about a "subject" for whom an Einstein or Dostoyevsky are only puppies - and we ants. We do not give much value to the intellectual ability of an ant, right? This gives a more accurate idea of the size of the danger. We're creating a possible "way of life" ultra-powerful who may or may not share our ethical and moral values. To discuss these implications, moreover, Kurzweil created an ambitious project: the Singularity University in Silicon Valley. It is an organization founded in 2008 that offers graduate courses focused on "assemble, educate and inspire a group of leaders." The fit these leaders "guide the development" of this artificial superintelligence what was to come. The hope behind the initiative is that these machines are born already "growing" values such as democracy, freedom of expression ... In short, make them, despite being vastly more powerful than us, not enough to have the "good sense" tear down. "People say,` Hey: it does not sound like a foolproof strategy, "says Kurzweil. "And is not it. But it's the best we can do."
If you can not beat them, join them. This saying will make more sense than ever in a world with machines billion times smarter than us: join they may be the future of humanity. "One day we can download our memories into a computer and preserve them," says another expert in intelligent systems: the Brazilian Miguel Nicolelis, who works in the development of prostheses able to converse with human brains - and function as if they were normal arms or legs . This "unloading", in theory, can mean uploading your consciousness into a machine. The mind still alive after the death of the body. And eventually lying eternally in the splendor of a simulated reality ... It is still a form of eternal life.
Well, probably it will be up to the machine to decide whether you will know that living in a simulation. She may find it better not to know anything, and go touch the lives thinking they have a body, you breathe, you will die one day ... If so, even the uniqueness may have already happened. And we would even now living in an illusion, a "Matrix". This hypothesis, at least philosophically, is not how to be refuted, since you can not imagine that a higher intelligence is really capable of.
Or have already done...
Assinar:
Postar comentários (Atom)


Nenhum comentário:
Postar um comentário