8 stories that affected the development of artificial intelligence
Today, not a single announcement about the release of a new smartphone or application is complete without a mention of artificial intelligence. AI in the trend, and you might think that he is with us quite recently, but it is not. Rather, it is not so: yes, AI instruments have been widely used today, but the development itself began more than half a century ago. And in the history of AI there were many interesting episodes not known to the general public, which we decided to tell the readers of Khabra.
Czech ), Immediately imagine a spoiler: first the robots worked for people, and then decided to prepare a riot that supposedly led to the disappearance of the human race. This idea does not seem new now. has become widespread in contemporary literature and cinema.
Officially, the term "artificial intelligence" was first used in 1956 by the young scientist at the Dartmouth Conference. John McCarthy . McCarthy explained AI as the science and technology of creating intelligent computer programs. Despite the differences in the interpretation of the terms of the final the judgment of , made following the meeting participants, was: "Any aspect of human intelligence can be accurately described in such a way that the machine can imitate it."
Already in the 20-ies of the XX century, artificial intelligence was not expected to be anything good. Images from the play by Karel Chapek RWR Source: Wikimedia
Episode 1: The mechanical turtles of Walter
Implement the concept of a robot with the beginnings of intelligence was able to American cybernetics Grey Walter. Its mechanical "turtles" , built in 1948-4? went to a light source and, resting on obstacles, handed over and drove around them. This was the first full-fledged manifestation of intelligence by the machine: instead of helplessly resting on the leg of the table, the robot concluded that it was impossible to travel and himself made the decision to maneuver a detour. In this case, the robot was created exclusively from analog components.
Gray-haired robot-tortoise Gray Walter on analog components. Source: extremenxt.com
Episode 2: The first machine-translator
In 195? IBM demonstrated an unfinished automatic translator from Russian to English, which operated only six rules and had a vocabulary of 250 words from the field of organic chemistry. Such a machine was required by the military to translate Soviet documents.
The demonstration made a furore in the press, which only spurred further funding in the field of AI. The benefits of potential automatic translators were obvious: it was estimated that out of ?000 full-time translators from different languages in the state service of the Joint Publication Research Service, only 300 people were downloaded per month. Improving the quality of recognition and automatic translation would give huge savings by reducing bloated staff.
The IBM translator from Russian to English, the future president of the United States Ronald Reagan and one of the fathers of computer science Herbert Grosh, 1954. Source: Columbia.edu
Episode 3: The first computer chess player
May 1? 1997 in New York, the computer for the first time in history won in the match, conducted by all the "human" rules, the current world chess champion. It is, of course, the second battle of Garry Kasparov and DeepBlue. It was 40 years after the programmer Service Bureau Corporation (division of IBM) and part-time chess player Alex Bersstein and colleagues wrote The Bernstein Chess Program - the first program for playing chess, working on the minimax algorithm on the mainframe IBM 704. The computer played slowly : for every turn it took him 8 minutes, for such a period of time he managed to calculate ?800 possible positions. Burstein himself never lost his own program, although he admitted: "Theoretically, 704 is not able to surprise me, but at times he does it - a couple of times he played so well that even confused me."
Mainframe IBM 70? playing chess at the amateur level
But the IBM 704 (not just this one, from the Service Bureau Corporation, and another one installed in the company Bell Labs) surprised really, so it's the writer Arthur Clark, who was in Bell Labs when the local craftsmen, having connected to the mainframe vocoder, taught 704 to perform the song Daisy Bell - for the sake of demonstrating the possibility of computer synthesis of speech. Writer Clark was impressed and put in the scene with the robot HAL 900? which performs the same song, in "Space Odyssey of 2001" .
But this is a completely different story. And in 195? the patriarchs of artificial intelligence, Herbert Simon and Allen Newell, predicted that the computer would be able to beat a person within 10 years. And indeed: in 196? the program Mac Hack began to successfully perform in tournaments along with people.
Episode 4: the brooding robot Shakey and the arrival of "AI Winter"
The development of computer technology and, in particular, AI solutions started in the 1950-60s, as shown above, very confidently, and initially led to impressive results. Moreover, the scientists began to make loud statements about the brilliant prospects for the development of technology. So, in 197? one of the fathers of the theory of AI Marvin Minsky predicted the emergence of computer intelligence, corresponding to the human, "for 3-8 years." With their loud statements about the near future AI scientists overstated the bar of expectations.
However, in the following years, scientists failed to make the promised breakthroughs - the main provisions and algorithms of artificial intelligence were developed, to the real products was far away, progress was extremely slow, the undeveloped area of artificial intelligence dispelled the efforts of researchers. One of the examples of unjustified expectations was robot Shakey , which was of great importance for science, but meaningless from the point of view of investors. Work on its creation and improvement was conducted in 1966-7? it was a mobile robot with a camera, which independently looked for a solution for the tasks assigned. Unfortunately, Shakey remained a test case for technologies and algorithms, without any utilitarian utility: it could function normally only in artificially constructed test space, and it took more than an hour to find a complete solution to the elementary task of "detecting a block on a raised platform and pushing it down".
The legendary robot Shakey with a very thoughtful artificial intelligence.
Source: SRI International
The growing skepticism of investors (and they were largely military) with respect to artificial intelligence reached its peak against the backdrop of a report by mathematician Sir James Lighthill, who said in 1973 that machines will always play chess at the level of an experienced amateur, and complex tasks like face recognition , they will never be able to do it. In addition to emotional reasoning, the report contained reasonable arguments against AI. Among them, the insufficient computing power of computers of that time and the lack of full-fledged machine learning were mentioned, because of which for the effective work of AI each time had to be filled with a huge amount of data.
Following the report, a sharp decline in the funding of research in the field of artificial intelligence marked the beginning of "AI-winter".
Episode 6: AI from Toshiba on the mail service (and not only)
Despite the cooling of investors' interest in the subject of artificial intelligence, engineers continued to work on narrow-profile machines with AI elements. They were not so ambitious as to try to create an "artificial mind", and came up with solutions to specific problems. But, unlike robots, the creators of such machines have achieved real success.
For the good half of the 20th century, all mail in the branches was sorted manually - operators read the indexes and addresses, shifting envelopes and postcards to the right cells. Attempts to create an automatic sorter began in the 1920s, but the first serial semi-automatic machines appeared only in the 1950s.
In the sixties, posters were used at the post office, reading indices on envelopes written strictly according to all known patterns. Deviation from this template made the index unreadable for automation. Especially if the index was written in the usual written handwriting, which a person can not always make out the first time.
It is these plain figures that are recognized even by the simplest sorting mail machines
In 196? under the leadership of the Japanese Ministry of Posts and Telecommunications, a project was launched to create an automatic sorting machine of a new generation. A year later, in Toshiba (then Tokyo Shibaura Electric Co.), a prototype of a mechanism recognizing printed figures written by hand was ready. And in 1967 Toshiba introduced a sorter with optical character recognition (OCR) technology. The machine scanned the envelope with a digital Visicon camera and sent the resulting impression to the recognition unit, where all unnecessary information was discarded, except for the digits grouped into the index. After recognizing the handwritten figures, the letter went to the desired sorting tray.
Post sorting machine Toshiba with advanced for 1967 AI. Source: Toshiba Science Museum
During the work of these first samples, the engineers assembled a database of 30?000 handwriting samples, greatly improving the effectiveness of AI, - as they say, they collected Big data, because the machine did not have the capacity for self-learning. And on July ? 196? the updated sorter, named Toshiba TR-? began work in the central post office of Tokyo.
Subsequently, developments were useful to engineers when creating the computer ASPET /7? which was a complex for the recognition of text from print media with a capacity of 2000 characters per second. If now OCR-scanner or OCR-application for smartphone does not seem surprising, then in 1971 ASPET /71 looked like a huge breakthrough.
Computer ASPET /7? which recognizes printed text.
Source: 電子 情報 通信 学会
During the second AI-boom of the mid-80's Toshiba introduced an improved version of the program-translator AS-TRANSAC , AI in which helped translate entire texts from Japanese to English and vice versa, preserving their meaning. Without the use of AI, which selected the desired values, the automatic translation would be an incoherent set of words.
In 199? the AI came in handy for Toshiba for the EUROPA framework, which allows creating voice systems that understand keywords in voice queries and respond to synthesized speech. Together with EUROPA, the company introduced MINOS - a platform for car navigation with recognition of voice commands and addresses.
Episode 7: AI in the 21st century
Really loudly artificial intelligence announced itself in 199? when the chess computer IBM Deep Blue beat Garry Kasparov, although a year before the machine lost to the chess player. In itself, the fact of losing a person to a computer was not something surprising, but the significance of the event was added to the hype in the media.
The release in 1999 of the home robocopics Sony Aibo inspired a false hope that the era of talking robots and all-powerful artificial intelligence is about to come, but the revolution did not happen again - after seven years Sony closed the project, restarting it only in 2017.
Prototypes of Aibo. By the way, in addition to dogs of robots, Sony produced lion and cosmonauts under this brand, although in much smaller quantities. Source: Alex /Flickr: DSC00193
A new boom of artificial intelligence occurred in the 2010s, when the power of computers and mobile devices allowed the use of AI in consumer devices and applications. For AI, all the stars finally came together: total digitalization helped create huge databases necessary for analyzing and learning AI, and instead of obsolete algorithms for learning neural networks, much more effective and productive ones were developed.
The IBM Watson Explorer AI system in 2017 successfully replaced 34 employees of the Japanese insurance company Fukoku Mutual Insurance, which analyzed the health of customers for individual insurance plans - the AI easily operates thousands of records, calculating the most beneficial insurance program for both parties.
The Google Brain algorithm. , working on YouTube, fills the sidebar of the recommended video, finding for each user the most interesting to him videos. Remember how many interesting videos you have discovered for yourself, thanks to YouTube recommendations.
The advent of AI on the trading floors gave a powerful push to e-commerce - the advisory AI on Amazon provides 35% of the total sales, valuing the scanned goods and selecting the products that they will buy with the most probability.
Artificial intelligence is already It is used in many creative mobile applications, from image generation (Prisma) to face masking (Snapchat), in all advisory systems, voice recognition systems, in most monitoring systems, smart homes, appliances, robots of all possible types, and so on. further, and so on.
And in the next decade the world will see available serial unmanned vehicles, the functioning of which is built on AI and machine learning. Last year, Toshiba introduced a new image recognition processor Visconti 4 , the first version of which appeared in the early 90's. AI in this chip is able to handle a huge amount of graphics data on the fly, accurately detecting video from cameras, marking, traffic lights, people, other cars and obstacles.
Episode 8: The machines learn to talk
Note: do not pronounce the recorded words like the IBM 704 in the eyes of the amazed Arthur C. Clark, namely to maintain a conversation with the person, responding adequately to his cues. The empirical test, proposed by the English mathematician Alan Turing in 195? was intended to answer the question that worried him at the time: Can a machine think like a man? Recall that the essence of the test lies in the blind communication of the judge with two subjects - a man and a machine. In the course of the dialogue, the thinking, in Turing's opinion, the computer must mislead the judge, masterfully posing as a human being.
The test found practical application in the competition for the Löbner Prize, held since 1990. Contestants present their chatbots posing as living interlocutors. By the way, the Löbner Prize, in addition to text communication, provides for a nomination in the audiovisual program, the victory of the AI in which automatically cancels the contest in the future, marking the advent of the era of complex AI. But so far in this nomination there has been no progress and hardly expected in the near future.
Here is an illustrative example that shows the difference between human intelligence and the fact that AI is called within the Löbner Prize (this is translated excerpt from the 2005 competitive program):
Dialogue between the judge and the person
Judge: Pzhschevic Parvishshch Krvahoch, right?
Member: Is it Polish?
Member: What you just wrote.
Judge: It's svsmrzya, is not it?
Member: I'm a little worried about your printing skills.
And now the dialogue of the judge and the winner of that chatbot Jabberwacky
Judge: OK, that's enough for now. I will communicate with another entity.
Member: OK, it was nice to use your comb.
Judge: What are you talking about? How did you use my comb?
Member: All right.
Judge: All right?
The participant: Yes
Judge: Very meditative conversation.
Participant: Moreover, I donated it to charity.
Judge: What? What did you donate to charity? This conversation?
Member: Um 1 million dollars?
Judge: What kind of charity?
Member: Are we talking about simpsons or unreal?
Judge: Honestly, I do not understand a damn thing at all what we are discussing here.
Since 2005 chatbots have made a big step forward. In 201? for the first time in the history of the Löbner Prize, the St. Petersburg bot Eugene (Zhenya Gustman), created by Vladimir Veselov, successfully deceived the judges in 33% of cases - according to the conditions of the competition, overcoming a 30% threshold means a successful passing of the test. That is, from 1990 to 201? despite the annual awarding of the most effective chatbot, the test for "humanity" artificial intelligence was formally not passed.
But until now the chatbots are far from the same artificial intelligence that fantasists dream about - the machine can not support the conversation with replicas that exactly correspond to the context. Human speech is incredibly complex and diverse and therefore still lends itself to complex computer analysis: understanding individual words and simple phrases, AI still does not understand the deep meanings, and in their cues, even with the correct arrangement of words, the computer's phrases seem absurd. And the most popular voice assistants Siri, Cortana and Alice from Yandex though sometimes give out very witty answers, but in any incomprehensible situation they stop the dialogue by switching to the search system.
Artificial intelligence is strong, but still far from perfect. The systems of prediction and recognition allow errors, although much less frequently than humans. While the society is talking about the ethics of the decisions of artificial intelligence, AI managed to take a new one, which seemed unattainable a couple of years ago: the AlphaGo program beat the man in Go, the last logical game where the primacy of the human intellect with its abstract thinking was considered unshakable.
It may be interesting