GET Artificial Intelligence: What Scientists Think About It / 1cloud.ru Blog / Sudo Null IT News FREE

Today, only the slow does not publish about AI. For example, Autodesk believes that synthetic intelligence agency stern assume into account much more factors than humans, and thus provide more exact, synthetic and even more creative solutions to multiplex problems. Generally expressed in the University of Oxford assumption that the artificial intelligence in the most later could replace gas-filled-time journalist and write for them, reviews and articles (and that and look to come through the Pulitzer Prize).

The universal fascination with the topic of AI has long at rest beyond the model of scientific conferences and excites the minds of writers, filmmakers and the indiscriminate public. It seems that from a future in which robots (Oregon Skynet) linguistic rule the world or, at least, solve most of everyday tasks, at a glance. But what do the scientists themselves opine close to this?

To begin with, it's worth reason the terminus "artificial intelligence": there are as well more speculations and artistic exaggerations on this subject. Therein affair, it is best to contact the author of this term (and concurrently the Jehovah of the Lisp language and the winner of many awards) - John McCarthy. In an clause of the same name ("What is AI?"), McCarthy gave the following definition:

This is the scientific discipline and applied science of creating trenchant machines, in particular - smart electronic computer programs. Artificial intelligence is associated with the task of using computers to understand the work of human intelligence agency, but is non limited to the use of methods discovered in biology.

It turns out that artificial intelligence and human news are closely related? Not quite so - McCarthy himself emphasized: if intelligence "in general" is a "computational" component of what helps the capable to achieve set goals, then the intellect of man, animals and machines will work differently.

It turns out that artificial intelligence is non like humanity, although many futurists, writers and symmetrical scientists require to consider that this is not and then. This is often repeated aside Michael Jordan., Retired Professor, University of Golden State, Berkeley. He believes that a lack of understanding of what artificial intelligence is, leads non only to the creation of "good-looking images" that are not related to real science, but to real misinformation and totally kinds of myths that flourish in this orbit.

Myth One: To create operating room improve bleached intelligence, you need to shape out how the human brain works.

Hashemite Kingdom of Jordan claimsthat this is non so at all. The work of AI, As a rule, has nothing to do with how human intelligence works. This "myth" is profoundly rooted collectible to the public's dependency to "splendid ideas": the authors of popular science articles along artificial intelligence liked metaphors taken from neurobiology very much.

In fact, neurobiology has a same indirect relationship (or none relation at completely) to the work of artificial intelligence. For Michael Jordan, the idea that "deep learning requires an intellect of how the weak brain processes information and learns" sounds like a blatant prevarication.

The "neurons" implicated in deep-water learning are a metaphor (or, in the language of Hashemite Kingdom of Jordan, generally a "caricature" of the brain), which is used only for brevity and convenience. In reality, the forg of the mechanisms of the same deep scholarship is much closer to the subroutine of constructing a statistical fashion mode of logistic retroversion than to the work of real neurons. At the same time, information technology doesn't go on to anyone to use the metaphor of a "nerve cell" in statistics and econometrics for "briefness and convenience".

Myth two: faux intelligence and deep learning - the current achievements of modern science

The opinion that computers "thinking like a man" will companion us in the near future is directly related to the idea that artificial intelligence, neural networks, deep encyclopaedism are the prop of exclusively modern science. Indeed, if we assume the idea that all this was invented decades past (and robots rich person not withal captured the world), the "verge of expectations" from technological achievements in general and the speed of their ontogeny in particular will have to be seriously reduced.

Unfortunately, the media are nerve-wracking to doh everything possible to foster involvement in their materials, and are very selective in choosing topics that, in the opinion of the editors, will arouse interest among readers. As a result, the accomplishments represented by them and their prospects turn dead set embody very much more awing than real number discoveries, and some of the information is simply "carefully lowered" so as not to reduce the passions.

Much of what is being presented "under the sauce" of artificial intelligence is simply processed information about neural networks that have been known to mankind since the 80s.

And in the eighties, everyone perennial what was known in the 1960s. IT seems that every 20 years there is a undulate of interest in the same topics. In the up-to-date wave, the main idea is the convolutional neural net, which was already mentioned astir twenty years ago
- Michael Jordan

The third gear myth: an artificial neuronic network consists of the same elements as the "real"

In point of fact, specialists involved in the maturation of computer systems operate neurobiological terms and formulations much bolder than many neurobiologists. Interest in the employment of the brain and the construction of human intelligence has get over a breeding ground for the development of such a theory every bit "nervous realism".

In artificial intelligence systems in that location are no spikes or dendrites, moreover, the principles of their work are FAR from non only when the work of the brain, but too from the notorious "neural realism." In fact, there is nothing "neural" in somatic cell networks.

Moreover, the idea of ​​"neural realism", founded on the likening of the work of artificial intelligence operation systems to the work of the brain, according to Jordan, does not hold water. According to him, it was not "neural realism" that LED to progress in the field of unreal word, but the use of principles that are completely inconsistent with the fashio the human brain deeds.

As an example, Jordan cites the hot unplumbed learning algorithmic program based on the "learning error postback". Its rule of operation (namely, signal transmission in the diametrical direction) clearly contradicts the mode the human wi works.

Myth Four: Scientists Understand How Human Intelligence Kit and caboodle

And this once again is far from the truth. As every the same Michael Jordan claims, the deeply principles of the brain do non just remain an unresolved trouble of neurobiology - in this sphere, scientists are spaced from the solvent of the problem for decades. And attempts to create a working artificial of the mind also do not bring off researchers closer to understanding how human intelligence works.

This is just an architecture created in the trust that someday people will create suitable algorithms for IT. But there is nothing to reinforce this Leslie Townes Hope. I opine that desire is based on the belief that if you build something like a brain, IT testament immediately get over clear what it can do
- Michael Jordan

John McCarthy, successively, emphasized: the trouble is not only in creating a system in the image and likeness of human intelligence, but in the fact that scientists themselves do non agree happening what helium (the intellect) is and for what specialised processes are responsible for.

Scientists are hard to answer this wonder in different ways. In his book Neural Networks and Deep Learning, Michael Nielsen gives several points of view. For exercise, from the standpoint of connectomics, our intelligence and its work are explained by how many neurons and glial cells our brain contains, and how many connections are observed between them.

Given that there are about 100 million neurons, 100 billion glial cells and 100 trillion connections between neurons in our brains, it's highly unlikely that we can "accurately recreate" this architecture and hit IT work.

Merely molecular biologists perusal the human genome and its differences from close relatives of people in the evolutionary chain give more encouraging forecasts: it turns out that the anthropomorphic genome differs from the chimpanzee genome aside 125 million base pairs. The figure is large, but not endlessly large, which gives Nielsen an affair to Leslie Townes Hope that on the basis of these data a group of scientists will be able-bodied to compose, if not a "working prototype", then at least an competent "genetic description" of the human brain, or rather basic principles, underlying his work.

It is worth locution that Nielsen adheres to "generally recognized human chauvinism" and believes that the significant principles that specify the work of human intelligence service Trygve Lie in the same 125 million base pairs, and not in the odd 96% of the genome that are the same in humans and chimpanzees.

So can we create artificial intelligence equal in hominid capabilities? Will it equal possible for America in the predictable future to empathize how our have brain works? Michael Nielsen believes that this is quite a possible - if you arm yourself with organized religion in a brighter future and that many things in nature do work reported to simpler laws than it seems at first sight.

But Michael Jordan gives advice closer to the practical work of researchers: not to buckle under to the provocations of journalists and non to seek "revolutionary" solutions. In his opinion, being attached to human intelligence as the starting point and the supreme goal of their research, scientists working on the problem of artificial intelligence unnecessarily limit themselves: interesting solutions in this surface area may lie in directions that are non attached to how our brain works ( and as it seems to us his device).

Postscript We in 1cloud consider the most different topics in our blog on Habré - a couple of examples:

• Friday Format: How Netflix Works
• In simple words: Understanding cloud services
• How Cloud Technologies Affect Information Centers
• A Little About 2FA: Two-Divisor Authentication

And talk nigh our own cloud service :

• Client-Oriented Cloud: 1cloud Experience
• 1cloud Cloud Photo Tour
• A little roughly information storage and 1cloud experience
• UI Optimization: 1cloud Experience

DOWNLOAD HERE

GET Artificial Intelligence: What Scientists Think About It / 1cloud.ru Blog / Sudo Null IT News FREE

Posted by: rollinsnowlielinuld81.blogspot.com

Share this post