How close is the war against the machines how to develop artificial intelligence the bobr times apple nutrition data

Those days when artificial intelligence (AI) was something from a terrible sci-fi movies, has sunk into oblivion. From the calculator in the phone and until an unmanned vehicle is all its different incarnations in the modern world. We found a way to teach computers in order to simplify your work.

In fact, every day we use it without realizing it. John mccarthy, who coined the term “artificial intelligence” in 1956, complained that since how the AI actually works, usually no one calls him that ( the neural network is also a manifestation of the AI). Because of this, the name is perceived as a kind of mythic structure, which still exists and which, vanninet only in the distant future.

Despite this, we still are, openly or subconsciously, afraid of artificial intelligence. Why? We’ve come to associate him with the movies he usually acts as an antagonist and something that tries to destroy people.Artificial intelligence

many people, when hearing about artificial intellect, imagine something like the war of humans and robots in “the matrix”. Or a sinister “skynet” from the terminator.

But this is not true. Until that AI is many, many processors. And they do not harm unauthorized person, and are at his service. What are the artificial “brains”

• narrow AI (UII), he’s weak. Specializiruetsya in one area. For example, can defeat the world chess champion. But it’s the only thing he can. If you ask him to choose the best way to store data on the hard drive — will make its figurative “blink”. His people are conquered and subjugated their needs.

• general AI (AIS). Sometimes called the strong, or the human level. Describes a computer that can perform operations in various fields, in fact, it looks human. Its creation requires more effort than UII, because he should be able to justify, plan, solve problems, to think abstractly, to comprehend complex ideological concepts to learn quickly and learn from experience with the same ease as we do with you.“rural stupid”

• artificial superintelligence (ISI). The oxford philosopher nick bostrom describes it as the intelligence that is smarter than any human in virtually every industry, including research and creative activity, general wisdom and social skills. Rank it can from the system that a slightly smarter person, so that in trillions of times ahead of any human being on all fronts.

Do not think that artificial intelligence will just come and take over the world, immediately, without a declaration of war. Actually, a narrow AI has long been used in all spheres of our life. He’s as smart as people, or exceeds our powers, but only in some narrow area.

The car is simply packed with such devices, because in modern cars the computer decides whether to submit fuel into the system if you, say, forgot to take off the handbrake. Not to mention the google self-driving car, which can respond to what is happening around him and analyze a situation, and can detect a pedestrian in a crosswalk before people notice him.Artificial intelligence

Phones – a small factory of artificial intelligence. For example, navigating city traffic, advisory services music and cinema, which are built on our tastes.

Google translate has significantly improved the translation quality through the use of a UII. Voice recognition is another task that requires the other system. But if you combine them, we get the possibility of simultaneous machine translations when you speak a sentence in one language and your partner receives it instant transfer.

Spam filters in email, product recommendations on amazon, or friends on facebook. All this is a vast AI network that mutually informs each other about who you are, what you are interested in, and analyzing the received information is your personal offer.

The choice of the gate at the airport for your plane airfare, search engines are able to process many different requests and generate the desired results, a system that helps doctors to diagnose, and even watson of IBM who once defeated colleagues anatoly wasserman in jeopardy, the equivalent of “your game” are too narrowly focused AI.Artificial intelligence

These systems are harmless. The worst that can happen — badly programmed UII could trigger an isolated local disaster, say, accidentally turning off the power supply, or bringing down the financial market. Similar has happened when in 2010, the artificial intelligence correctly reacted to an unexpected situation and provoked the withdrawal of a trillion dollars out of the market). While AI is not capable of threatening our existence, we are quietly yellow bricks lay the way to more perfect types, until sooner or later wake up something. From private to general

Nothing can make so much respect for the power of human intelligence, as scientists attempt to recreate it in artificial form. To build skyscrapers and space ships is proving to be easier than to recreate or even understand your own brain. This is the most complex object in the known universe.

The most difficult part was to create an intelligence as smart as people, as a whole, not in any particular region.Artificial intelligence to create a computer that will multiply the number with twenty zeros, or to distinguish cats from dogs easy, but to get to understand the content of picture books-no.

And what we think it is just basic and does not require many resources, in fact millions of years perfected in us and our evolutionary ancestors. We do not think about how our body, when we raise the hand in our brain is the perfect software that performs this operation. All “captchas” on websites work on the same principle as our brain can understand the strange interweaving of meaningless letters, and for AI such manipulations — as long as unattainable value.

Everything already knows how narrow the intelligence, new to humanity, it is unlikely that our ancestors ever played chess or go, and have multiplied trillions on trilliardy.

However, in many areas of descriptive action, people can easily bypass the computer due to the fact that for thousands of years to recognize facial expression, the shades, the structure, the differences between two-dimensional and three-dimensional space.Artificial intelligence that machine will only set black-white-gray tones, for the man will develop into a three-dimensional picture of the meteorite. And yes, it is unlikely that the computer will be able to distinguish why the “brave heart” well, but “patriot” is worse.

Scientists have a plan to equalize the car and the man. The first is to increase the capacity. At least, in order to achieve human abilities, we need more computational capacity. One of the ways of expression is the total number of computations per second (cps). Ray kurzweil, google’s chief futurist, empirically came to cps to person — 10 to 16 degrees, that is 10 quadrillion calculations per second.

Currently the world’s fastest chinese supercomputer tianhe-2 has broken this figure, the record was 34 quadrillion. But it is a huge edifice, occupying 720 square meters (n in almost 10 football fields), which consumes 24 MW of electricity, which is comparable to a small city, while our brain is only 20 watts, and is a chinese treasure $ 390 million.Artificial intelligence not so for wide, even for commercial use.

To determine the speed of approaching education with advanced artificial intelligence offered another amount. The more cps you can buy for $1000, the closer humanity to the point of transition to general AI. And if you start from the moore’s law, which says that every year computing power in the world doubles, and now all the computers in the world are at the level of the rat brain. Doesn’t sound too impressive, until we remember that in 1985 it was a trillionth fraction of human-level billion in 1995 and million in 2005.

The second part of the plan is to make the AI smarter. Frankly, nobody knows exactly how to achieve this. While there are several assumptions. For example, to copy, to put it simply — to make plagiarism of the brain. In fact, the current supercomputer is a prototype of our brain, only a very long time. Scientists are actively working to study the development of our brain, to understand how it evolved to its current state.“the matrix” according to optimistic estimates, the study will be finished by 2030, and it can inspire engineers. An example of such synergy — neural networks that learn from their experience, with each new task building new relationships and optimizing themselves. Our brain learns in a similar way.

There is another, equally interesting approach is to emulate the mind – scan of the brain is cut into thin layers, layer by layer, to create a real three-dimensional model and implement it in a computer. Get a computer with the capacity of the brain, which can collect information and learn.

If the engineers succeed, the brain that dies as a jim, wakes up the same jim. That is, will be a real shift of personality in your computer.

Until scientists were able only to emulate the brain of a flat worm, just a millimeter in length, consisting of 302 neurons. A human has 100 billion. Does not look very realistic, but things can change.Artificial intelligence

But while it is impossible to copy the brain, you can try to copy it the evolution of the method of “genetic algorithms”. So was the selection of living beings – they lived and either multiply, passing on the best genes or not. In computer evolution can engage a group of computers that will perform such an operation, the best examples of their software code can be mixed and output to the new system, the company level, when this process will happen by itself. But this process in humans took millions of years, we want the result for a couple of decades.

However, evolution also has its peculiarities. First – completely useless branches. Second — she did not have a specific purpose, for example, to create a superintelligence (and sometimes even selected individuals, they are not shining). Third — for its development it needed modifications, for example, ways of using energy in the body. So while it is not clear whether to elect an evolutionary strategy.Artificial intelligence

The third way is not our problem, and the computer let him decide. Oddly enough, the most encouraging method. The idea is to build a computer that will study the AI and he will make changes to the code, improving the structure and making themselves more reasonable. Based on these assumptions and the growth in capacity exponentially new level of AI can appear very suddenly. A superintelligence

Computers already surpass humans. The maximum frequency of the human neuron is 200 hz, and the computers exceeded 2 ghz — they are 10 million faster. Signals in the brain can move at a speed of 120 m/s, whereas in the car — at the speed of light. And while the brain is forever encased in the skull, computers can expand to any value, and the transistors in circuit boards is not a problem to recover, in contrast to nerve cells, working 24/7 without rest. In addition to the “zhelezyachnye” part, computers can life improve through updates and patches, leaving a wide field for experimentation.Artificial intelligence and even then, what people surpassed all other types — creating collective intelligence for computers is unlikely to be a problem.

And if AI will be able to develop itself, reaches the point of human intelligence, he will not perceive as a kind of milestone, it seems it is only us. After all, he had no reason to stop there. We think he will develop for a long time, and even when he reaches our level (by nick bostrom, “rural stupid”), we will still be funny. Only between “rural stupid,” and einstein, by his standards, nothing at all.

Scary? Once the system reaches the human level, it will need too much time to reach the level of intelligence of all mankind. According to realistic forecasts, this can happen in 2045, according to pessimistic — in 2075-M. After that, there will be an explosion of intelligence, and it quite rapidly reaches the level of ISI, which we can’t measure — if in our understanding of IQ “rural stupid” 80 wiseacre — 130, the value in 12952, we can not be aware of.“rural stupid” to increase this potential can as in a few years, and, say, 90 minutes.

Imagine that in the end we get a “good” AI, and the scripts of “the matrix” and “terminator” are deposited. In the end it will be something from the “hitchhiker’s guide to the galaxy”, which can be operated in three models:

• “sovereign” that can manage the processes around the world and free to take decisions independently, for example, as people to move better.