Portal field news

Portal field news

in ,

🌐 | The Internet of Things (IoT) market recorded an average annual growth rate of 6.1%, and in 2020 it will be 83 billion US dollars ...


The Internet of Things (IoT) market recorded an average annual growth rate of 6.1%, and in 2020 it will be 83 billion US dollars ...

If you write the contents roughly
Artificial intelligence, data analysis, and machine learning play important roles in industry.

In 2021, the global Internet of Things (IoT) market was worth US $ 48 billion.Internet of things in the world ... → Continue reading


Atpress is a pioneer in the distribution of press releases.
We handle a wide range of genres of news, from tourism and gourmet to entertainment and gadgets.
It is packed with the latest information that can be delivered faster than anywhere else because @Press, which distributes press releases to TV, magazines, web media, etc. and is strong in SNS!

Wikipedia related words

If there is no explanation, there is no corresponding item on Wikipedia.

Artificial intelligence

Artificial intelligence(Jinkou Chino) orArtificial intelligence(British: artificial intelligence,AI[AI]) means ""Calculation(computation)” and “Computer(computer)”IntelligenceTo studyComputer science(computer science) A field”[1]. "言语 OfUnderstanding,inference,problem solvingIntellectualcommunication skillTheA humanLet the computer do it for youTechnology'[2], OrcomputerIntelligent by (computer)Information processing system OfdesignAnd regarding realizationResearchAlso called "field"[3].

"Complete Encyclopedia of Japan(Nipponica) ”,Information engineeringSato, a communication engineer, said as follows:[1].

"If you don't be afraid of misunderstanding and simply change your mind, you can say,recognition, Inference, language operation, creativity, etc.), what procedure (algorithm) And whatデータIs it a field to study "whether (prepared information and knowledge) can be prepared mechanically?"[1].'


Various technologies that realize human intellectual ability on a computersoftware-Computer system[4]..As an application exampleNatural language processing(Machine translation-Kana-Kanji conversion-Parsing・ Sentence summary, etc.)[5][6], Imitate expert reasoning/judgmentExpert system, Analyze image data and identifypatternDetect and extractImage recognitionEtc.[4]. In 1956Dartmouth ConferenceJohn McCarthyWas named by. Currently, the description of intelligence usingInformation processingIt is also used in the sense of approach in research.Household electrical machinery OfControl system,Game software OfThinking routineIt is also called this way.

Programming language LISP by"ELIZA"Our TeamImitatedProgram(Artificial brainless) Is often referred to, but in the realization of a research / information system called an "expert system" that attempts to make a computer play the role of a human expert, the description of common knowledge that humans implicitly has becomes a problem.It is considered difficult to use it for practical purposes.[When?]As an approach to the realization of artificial intelligence, "Fuzzy theory"Or"neural network, Etc. are also known, but it is conventional artificial intelligence.GOFAI The difference from (Good Old Fashioned AI) lies in the symbolic clarity of the description. afterwards"Support vector machineAttracted attention. Also, learn based on your own experienceReinforcement learningThere is also a method. "In this universe, intelligence is the most powerful trait (Ray KurzweilAs is the case with the word ")", mechanically expressing and implementing intelligence is an extremely important task.

In image processingDeep learningRecognized worldwide for its usefulness2012 Research became active rapidly from around that time, and the third artificial intelligence boom arrived.[7].. From 2016 to 2017, AI that introduced deep learningFull information gameIsGoTop players such as, and even an incomplete information gamepokerThe world's top-class players[8][9],MahjongThen "Microsoft Suphx (Super Phoenix)" reaches the first stage as AI for AI[10]Is attracting attention as a state-of-the-art technology[11].

Science and engineering overview

To take advantage of deep learningLinear algebra,Probability theory,statisticsMathematics above university level, such asData scienceKnowledge of (data science) is required.

To perform a brain simulationNeuroscience, Social analysis also requires knowledge of applied fields such as sociology.

さ ら にImplementationto do soprogrammingKnowledge is also required.The programming language isC++BesidesPythonIs widely used.

AI type

Artificial intelligence in the second artificial intelligence boomMachine learningThere is something like the following.

Expert system
A conclusion is obtained by applying the inference function. Expert systems can process large amounts of known information and provide conclusions based on them. For example, in the pastMicrosoft OfficeHas a system that when a user types in a character string, the system recognizes certain features and makes suggestions accordingly.
Case-based reasoning (CBR)
Based on a past case similar to the case, a partial correction is made and a trial is performed, and the result and the case are stored in the case base.
Bayesian network
Behavior-based AI: A method to build an AI system from scratch

On the other hand, computational intelligence (CI) is based on repeating development and learning (for example, parameter adjustment,ConnectionismSystem). Learning is an experiential method, with non-symbolic AI and unaesthetic AI[Annotation 1],Soft computingIs related to The method is as follows.

neural network
Pattern recognitionIt is an algorithm specialized for.Almost synonymous with connectionism.
Fuzzy control
It is a technique that enables inference in uncertain situations by means of sets defined to tolerate ambiguity, and has been widely adopted in control systems since the 1990s.
Evolutionary computation
It is a technique inspired by biology and seeks the optimal solution of a problem by applying the concepts of evolution and mutation. This methodGenetic algorithmSwarm intelligenceare categorized.

Attempts have been made to create an intelligent system that integrates these.ACT-RNow, let's apply the inference rules of experts to neural networks andProductionGenerate through.

1990 era, Neuro-fuzzy home appliances became popular in Japan, but these home appliances are products that apply neural networks and fuzzy control researched in the second artificial intelligence boom.

In the second artificial intelligence boom, disappointment with artificial intelligence spread because it was only a subtle improvement of existing products and did not lead to products that enable the high degree of automation that people envision.

In the 3rd artificial intelligence boom, deep learning has achieved much higher accuracy than the artificial intelligence of the 2nd artificial intelligence boom in various fields such as image recognition, text analysis, and voice recognition, and research on deep learning is active. It is done.recently,DQN,CNN,RNN,GANAnd various derivations of deep learning have emerged and are active in various fields. In particular, GAN (Adversarial Generation Network), in addition to deep learning's success in areas such as recognition and prediction, has shown significant advances in image generation technology.Masaya MoriPoints out that the application field of conventional artificial intelligence is expanding against the background of these results, and the application of creating content called Creative AI has also started.[12]..Similar to the former neuro-fuzzy home appliances, many home appliances that claim to be AI-compatible have been released, and deep learning is also applied to these.

Global case study of AI research and development

President of the Times in 2013 in AmericaBarack ObamaIs a brain research projectBRAIN Initiative"Announced.

GoogleIs working with the Allen Institute for Brain Science to develop software for processing large amounts of data generated by brain scans. As of 2016, the amount of Brainmap data managed by Google is already 1.ZettabyteHas reached[13][14].. Google has also begun collaborative research with the Max Planck Institute in Germany, conducting research to reconstruct neural circuits from electron micrographs of the brain.[15].

2016th 13 in China5 year planAI as a national project[16], As a brain research project(English editionAlso launched[17], The public and private sectors are promoting AI research and development[18].. Chinese educational institutions have also gathered genius children under the age of 18 and openly invested in the development of AI weapons[19].Massachusetts Institute of Technology(MIT)(English editionProfessor(English editionAccording toPoliciesIt is said that it is in an environment where it is easier to research AI and experiment with new technologies compared to the conscious Western countries[20][21][22].. Motoaki Saito, who promotes research and development of supercomputers in Japan, also asserts that China may lead in the development of AI[23].. It is said that China occupies three-quarters of the world's deep learning computers[24].. According to the US government, since 2013, China has surpassed the United States in the number of papers on deep learning since XNUMX.[25].(English edition,(English editionThe Chinese are monopolizing the top ranks in AI global competitions.[26][27].. Major AI company Google,Microsoft,AppleWas also an executiveTaiwanese AmericanScientistKai-fu LeeSay that China is taking over the hegemony with AI.(English editionWas published by the American political world and media.[28][29].

French PresidentEmmanuel MacronDeclares to spend $ 5 billion over five years to support development in the AI ​​space[30], Opened AI Institute in Paris and invited Facebook, Google, Samsung, DeepMind, Fujitsu and others.Long-term collaboration in AI research has also been decided with the United Kingdom. The EU as a whole will invest 2020 billion euros through the "Horizon 215" plan.South Korea will invest $ 20 billion by 2022. Six AI institutions were established and a reward system was created.The goal is to be in the top 6 AI world by 2022[31].

According to a survey by the Nikkei newspaper, the number of AI research treatises by country was 1st in the United States, 2nd in China, 3rd in India, and 7th in Japan.[32].

Cultural and artistic application examples

In the music field, by learning existing songs, you can compose by imitating the style of a specific composer.Automatic compositionSoftware has appeared.AlsoRhythm gameA system specialized in the field has also been developed, such as automatically generating a musical score showing the touch position used for music from music.[33].

In the field of painting, the background and animation for concept artMiddle discountAutomatically generated, monochromeComicAI that assists human work, such as automatic coloring of[34][35][36].

Shogi AI is learning from human-to-AI games to create new tactics, but it is useful to actually point to it, although it is incomprehensible to professional shogi players (humans).[37].

Discussion / social issues

Artificial Intelligence Society OfYutaka MatsuoIn his book "Does Artificial Intelligence Exceed Humans?"A humanAgainst反 乱Denies the possibility of causing.Itsuki Noda, the chairman of the Japanese Society for Artificial Intelligence, said, "Cingularity (technical singularity)"Or"AI surpasses human intelligence in 2045"It's a science fiction story," he criticized.[38]..He said:[38].

in the first place"IntelligenceThere is no point in talking about what "" means without defining it.
If intelligenceCalculationIf it points to power80 eraIt will be beyond the ability of human beings[38].

Sociologist Robert M. Gerachi says with AIRoboticsMajoring in (robotics)Carnegie Mellon UniversityAs a result of a field survey of the institute, it was concluded that the actual research was "the mundane reality", which was far from the singularity theory.[39]..He is KurzweilHans MorabeckThe theory of singularityEschatologyWe call it "Apocalyptic AI" and assume that we do not support or refute such a theory itself.[40]..On top of that, "Apocalyptic AI"Popular scienceConsidered as a kind of (pop science)[41]..So it ’sReligious-Entertainment-fictionUsing easy-to-understand and intriguing explanations similar to, etc., attracting a large number of interests and earning research funds[41]..Gerachi

Apocalyptic AI is, in fact, a demand for money. ("Apocalyptic AI is, indeed, a request for money.")

It has said[41].

On the other hand, there are also ideas and claims that regard artificial intelligence as dangerous.

Human rights violations

Massachusetts Institute of TechnologyProfessor of(English editionWith huge financial powerhuman rightsHas the repression ofChugokuSucceeds in the competition for artificial intelligence development, saying that the preconceived notion that democratic nations dominate technological innovation will change.[20]Called one of the "Fathers of Deep Learning"Yoshua BengioWarns that China is using artificial intelligence for civilian surveillance and political purposes.[45][46], The world's human rights groups and the media have called "digital authoritarianism" a political system that suppresses human rights with artificial intelligence represented by China.[47][48]"Digital dictatorship"[49][50][51]"DigitalPolice state'[52]"Digital totalitarianism"[53]"AI dictatorship"[54]I called.

In China, a project supported by the Chinese government is being promoted to monitor the brain waves and emotions of the people with artificial intelligence from sensors embedded in helmets and hats.[55][56][57][58],Online censorship[59][60]Officials,刑 務 所From prisonerscrosswalkArtificial intelligence to monitor even pedestrians[61][62][63][64][65][66],Surveillance cameraPolicemen OfSunglassesSmart glasses[67],Mech robot[68]ToFace recognition system(Sky net) With artificial intelligenceSurveillance society-Management societyHas been made[69][70][71][72].

Xinjiang Uygur Autonomous RegionThen, surveillance camerasMobile phoneAnalyze personal information collected from such as with artificial intelligence(English edition,Race profilingOf the ethnic groups selected inUighurAs of June 2017, about 6 people did not go through legal procedures.terrorism,crimeAs it may violateXinjiang Uygur Re-education CampToPreventive detentionIt is an internal document of the Chinese government(English editionHas been reported[73][74][75], Government-based selection of specific ethnic groups using AI, and computers to humansConcentration campSend toHuman rights violationHas become an international issue as "unprecedented"[76][77].Hong KongThen, the fear of becoming a surveillance society by artificial intelligence similar to mainland China[78],2019-2020 Hong Kong Democratization DemoWhen it happensSurveillance cameraSmart equipped withStreet lampHowever, it was destroyed one after another by the citizens[79][80].

China uses AI surveillance technologyMiddle East-Asia-AfricaAnd export to all over the world[81][82][83][84][85][47],United Nations OfSpecialized institutionIsInternational Telecommunication UnionHuman rights groups are concerned about the spread of human rights abuses like China, as China is also leading the international standardization of AI surveillance technology through the (ITU).[86][87].

ChineseSocial credit systemThere is a concern that a system that uses artificial intelligence to utilize big data to determine the aptitude of people, as represented by, will lead to fixing the gap between social classes.[88],European UnionThen, from May 2018, in employment and financing only by big data analysis of artificial intelligenceDiscriminationDo not admitEU General Data Protection RegulationWas enforced[89].

Massachusetts Institute of TechnologyFace recognition systemWith the accuracy ofMicrosoftAnd China's Megvii are over 9%,IBMReached 8%Amazon.comIs 6%RacismThere was a dispute with Amazon.com when it published a study that said it had a positive bias.[90].

The semi-automatic quadruped robot "Spot"[91],New York City PoliceWas deployed on-site by the citizens, but was discontinued due to public protests[92].. "Generally, most Spots are used to investigate broken power lines and gas leaks." On the other hand, police purchased and used Spots without the approval of the citizens, so they are "robot dogs." Was labeled[92].

Military use

Armed forces of major powers are trying to automate in the field of missile defense. The US Navy has a fully automatic air defense systemPhalanx CIWSCan be used to destroy anti-ship missiles with a Gatling gun. Israeli army anti-aircraft missile systemIron domeOwns theGaza StripThe target is automatically detected on the boundary line betweenGuardium,Samsung RCWSIs running and shooting multiple humans[93][94].. It is also argued that AI will create new military capabilities, change the command of the military, training, deployment of troops, transform war, and that change will determine the military balance between large powers.[95].P-1 (Patrol aircraft)It may be installed in the battle command system for support like.

June 2016, USUniversity of Cincinnati"ALPHA", which was developed by a research team of the company, was announced to have unilaterally won a simulated air battle with a former US military pilot. AI programs use genetic algorithms and fuzzy control, which does not require high processing power to operate the algorithms.Raspberry PiCan work on[96][97].United States Department of DefenseIssued a ban on the development of autonomous killing weapons that did not rely on human judgment from a humanitarian perspective in 2012, and made it permanent in 2017.[98].

Some scientists and tech leaders argue that military use of AI will accelerate global instability. In 2015ブ エ ノ ス ア イ レ スHeld in Stephen Hawking, an American space venture companySpace xFounder Elon Musk,AppleCo-founder ofSteve Wozniak, An open letter was issued by scientists and entrepreneurs.DroneAI-equipped weapons such as humanoid robots that operate firearmsGunpowder,nuclear weaponsIt is seen as the third revolution that follows, and some of them are predicted to be practical within a few years.It was used for national instability, assassination, oppression, and selective attacks on specific ethnic groups, and it was clarified that the competition for weapons development would not be beneficial to humankind.

In March 2015,Harvard UniversityLaw school and international human rights organizationHuman Rights WatchDemands a ban on autopilot weapons[99].. In March 2017,United NationsFirst Official Expert Meeting on Military Use of AI Held in[100], In August 2019, the conference adopted virtually the first international rule on the operation of AI weapons, but it was not legally binding.[101].

East-West conflict

New Cold War,US-China Cold WarThe United States, China, and Russia, which are also said to be in a state of affairs, are competing for development comparable to nuclear development over the military use of artificial intelligence.[102].

China has 2017 aircraft in June 6DroneIn the group's autonomous flight experiment, the record of the U.S. military that succeeded in the flight experiment of 2016 aircraft in 103 was set, and in May 2018 the following yearNorth AmericaAlso announced a CG image of bombing the city of[103]In June of the same year, the world's largest test using 6 autonomous unmanned boats[104]AI military technology (especially a large amount called swarm)Loitering weaponsIt is also argued that the United States needs to prepare for the future, fearing that China is making rapid progress in the integrated operation of autonomous weapons such as[95].

China's military AI developmentU.S. ForcesIn March 2019, giving a sense of crisis to the political worldJoseph DunfordChairman of Joint Chiefs of Staff,Patrick ShanahanSecretary of DefenseActing,Donald TrumpPresidentWith the establishment of an AI research base in China, etc.Chinese PLAWhen cooperating withGoogleBlame[105][106], Google CEOThunder PichaiInterviewed with Dunford and President Trump and explained that the results of the AI ​​research center in the People's Republic of China are open to all people, not just China.[107].

In the United States, it has been revealed to Google employees that Google was conducting a top secret plan "Maven Plan" to cooperate with the military use of AI by the U.S. military.[108], March 2018United States CongressAt the hearing, a secret plan to cooperate with the similarly exposed Chinese government(English edition”, And was pursued in line with the six principles of artificial intelligence development in June 2018, which Google vowed to refuse weapons development and human rights violations using artificial intelligence.[109].

PLA fighterJ-20Target selection supportalgorithmWhen it was reported that Google's AI researchers were involved, he denied that it was "statistical modeling, not AI."[110]..It also caused ripples when Microsoft announced joint research with Chinese military educational institutions and AI.[111].

January 2019Mark EsperThe Secretary of Defense said that the People's Republic of China is not only building a new surveillance nation with AI, but also in the Middle East.Pterosaurs,rainbowな どDroneWarned that it also sells drone weapons that autonomously attack with AI by spreading a large amount of[112].

Russia and China are said to have already been put to practical usehackingOther automation[113]Attack a specific individual,Deep fakeSpoofing and bot postings raise concerns about controlling public opinion, etc.[114].

Employment problem

Oxford UniversityDr. Michael Osbourne of2013 According to a paper published in, about 49% of the working population in Japan is highly likely to be replaced by artificial intelligence and robots (47% in the United States and 35% in the United Kingdom).2030 erauntilfast foodEmployees who cook at the storeMech robotIt has been pointed out that there is a high probability that it will be replaced by AI and AI with 81%[115][116][117].

However, for this treatiseLaboratoryThere are criticisms that it is overestimated because it includes jobs that can be automated at the level. When Michael Osbourne came to Japan in October 2016,Institute of Economy, Trade and IndustryKoichi Iwamoto asked, "What kind of intention and assumption did you make the calculation?" ing.This exceeds humansShogiBecause AI is emergingShogi playerHowever, it merely showed that it could be replaced by artificial intelligence, and did not respond to the desire to watch games between humans.Today's professional shogi players use AI to study opponents and situations, as well aselmo enclosureThere is a coexistence relationship such as using the tactics that AI first used[37].

And make up the professiontaskStudies have shown that when viewed on a (business) basis, only about 70% of occupations (about 9% in Japan) have tasks that exceed 7% automated.There are also claims that AI and mechanization will deprive employment, but those technologies will reduce the amount of tasks, but jobs that introduce and maintain AI and mechanization, and those technologies will create new jobs. This can also create jobs.At the same time, however, the number of routine jobs with moderate skills has decreased, and the work has been polarized into low-skilled jobs and highly-skilled jobs that do not require specialized skills.Economic disparityIs also expected to expand[118][119][120].


Although the construction of AI has been attempted for a long time, solving the symbol grounding problem and the frame problem has become a major obstacle.


Early 17th century,Rene DescartesProposed that the body of an animal is just a complex machine (Mechanics).Blaise PascalFirst in 1642Mechanical calculatorWas produced.Charles BabbageAda lovelessHas developed a programmable mechanical computer.

Bertrand RussellAlfred North WhiteheadPublished The Principles of Mathematics and revolutionized formal logic.Warren McCullochWalter PittsPublished a paper titled "Logical Calculation of Ideas Implicit in Neural Activity" in 1943, laying the foundation for neural networks.

Late 1900s

In the 1950s, AI was beginning to produce vigorous results. John McCarthy at the first conference on AI[Annotation 2]Was created. He is also a programming languageLISPWas developed. As a method to enable testing regarding intellectual behavior,Alan TuringIs "Turing testWas introduced.Joseph Weisenbaum TheELIZAWas built. this isVisitor-centered therapyI do[Annotation 3].

In the proposal for the Dartmouth Conference held in 1956, it was first used as a term in human history and was established as a new field.

Between the 1960s and 1970s,Joel Moses The Macsyma(MaximaProgram[Annotation 4]In it we show the power of symbolic reasoning in integration problems.Marvin MinskySeymour PapertPublishes "Perceptron" to show the limits of simple neural networks, Alan Carmelauer is a programming language Prolog Was developed. Ted Shortriffe established a rule-based system for medical diagnosis and therapy,Knowledge representationAnd showed the power of reasoning. This is sometimes called the first expert system.Hans MorabeckHas developed the first computer-controlled vehicle that runs autonomously on a course with obstacles.

In the 1980s, neural networksBack propagationWidely used by algorithms.

Also in this eraRodney brooksHowever, he advocated the theory that the body is essential for intelligence (physical nature).

In the 1990s, various applications were successful in many areas of AI. Especially in board games, in 1992IBMDevelops TD gammon, a computer dedicated to backgammon that is comparable to the world champion,deep blueIn August 1997Garri KasparovWas beaten by Othello in August of the same year.NECOf Othello dedicated computer · World ChampionKen MurakamiLost[121].Defense Advanced Research Projects BureauIs the firstGulf WarHe used AI to schedule units at and found that the cost saved by it outweighed all government investment in AI research since the 1950s. In Japan, Shunichi Amari (member of the Japanese Academy) enlightened energeticly and produced excellent results, but the black box nature of logic was pointed out.

1998 ToUnstructured dataFormalInternational standardIsXMLWas proposed, but from hereWebAn attempt has been made to apply the semantics suitable for each application to the above unstructured data and to perform processing. In the same year,W3C OfTim Berners-LeeAllows the Web to perform intelligent processingSemantic WebWas proposed. This technology adds a meaning to the data on the Web and internationally standardizes a method for causing a computer to perform intelligent processing. This standard includes a data format that represents an ontology in knowledge engineering.OWLIt was once popular because it also includedExpert systemIt turns out that it is a variant of. Although the standardization was completed in the first half of the 2000s, it has not been popularized yet because the Web developers could not find the merit corresponding to the development man-hours.

Second AI boom in Japan

In Japan, after the trend of expert systems, neuro-fuzzy became popular. However, as the research progressed, there was no shortage of computing resources and data, symbol grounding problems, and frame problems, and there was no AI that would radically change the way the industry was, and the boom ended.

Expert system (application of knowledge engineering)

In the 1980s, a large number of expert systems based on knowledge engineering came to be proposed, mainly in laboratories of large corporations, and AI ventures specialized in expert systems were launched one after another. The fifth generation computer is the ultimate project born from the fashion.

Fifth generation computer (high-performance Prolog inference machine)

Japan spent 1982 billion yen as a national project from 1992 to 570Fifth generation computerHowever, the knowledge engineering method adopted requires enormous manual input of rules, and there is a problem that unified rules cannot be made when the interpretation of expert knowledge differs among experts. The real expert system was not realized. The actual product is a parallel type Prolog dedicated machine that directly interprets Prolog instructions by the CPU hardware mechanism and executes them at high speed, but in the commercial sense, no application destination was found.

Neuro Fuzzy[122]

From the late 1980s to the mid-1990s, ON/OFF control that has been used as a conventional electronic control method,PID control,Intellectual control is actively studied and knowledge engineering rules are used to overcome the problems of modern controlFuzzy control,Learning and classifying data featuresneural network, A method called neurofuzzy that combines the two has reached a boom mainly in Japan. In 2Subway opened in Sendai OfATOAdopted for[123]Then, in line with the high-end routes during the bubble era, many models of white goods, which have significantly increased the number and types of sensors and optimized driving based on various data, have begun to be released.

As for fuzzy, it is known that Japan had a particularly big boom in Japan, as Japan has obtained 2018/1 of the patents of the world by 5.[124].. Today's white goods use control technology that has been further developed since that time, but it has already become commonplace, and users are not aware of it. Neuro-fuzzy has become a boom1990 eraStillbig dataThere is no concept (after the spread of broadband connections2010 Was first proposed toData MiningWas not applied to industry. But,neural networkThis is the first case of an epidemic involving ordinary people,Deep learningIt can be said that this is a social phenomenon that can be said to be the prehistory of the boom.

History of the boom

Matsushita Electric started research on fuzzy control that makes use of human ambiguity in control from around 1985, and launched fuzzy washing machine No. 1990 "Aizuma Go Fuzzy" on February 2, 1. I rowed. "Aizuma-go Day Fuzzy" is a washing machine that optimizes driving flexibly based on data collected by more sensors than before, and was the first washing machine of its kind in the world. Since the introduction of the most advanced technology of fuzzy control at the time matched the high-class route during the bubble era, fuzzy attracted a great deal of attention in the world despite the fact that it was a control technology behind the scenes.[124].. The degree of fashion was such that "fuzzy" was selected in the gold medal in the new word category of the 1990 New Word/Buzzword Awards. After that, Matsushita Electric developed a neuro-fuzzy control that automates the complicated tuning of fuzzy rules, and was not only evaluated at academic societies by overcoming the limitations of conventional fuzzy theory, but also successfully applied to white goods. It caused a further boom. Following the success of Matsushita's efforts, other companies also launched a number of products that use similar intelligent control. Until the mid-1990s, the name of intelligent control technology was widely used as a sales phrase for white goods for general use by manufacturers, and the product name of a washing machine was "Aizuma-go FUZZY", a classification of vacuum cleaners. "Neuro-fuzzy vacuum cleaner" and "Neuro-auto" in the operation mode of the air conditioner[125][126][127][128][129][130].

The methods of neuro, fuzzy, and neuro-fuzzy require conventional simple on/off control and objective modeling of an object with mathematical expressions (this task becomes extremely difficult when the object has a complicated mechanism). Compared with PID control and modern control, fuzzy and neuro, which can use the characteristics of human subjective experience and measured data, can increase the flexibility when adapting to the environment while reducing the development man-hours. Had an advantage[122].. However, despite the efforts of the developers, due to the small calculation capacity and the small amount of data that can be collected, the control of existing machine tools and home appliances has reached the limit to some extent. Theoretically, this is a combination of fuzzy sets and neural networks incapable of deep learning, and there is a limit to the improvement of recognition accuracy even if abundant computation resources and data are given.

Since then, the theory has not improved steadily due to the limitations of the computer's ability, and there has been no noticeable progress. Boom has gone[131].. After the boom, it was not generally noticed, but as a behind-the-scene technology, it is used not only for home appliances but also for social infrastructure such as rainwater drainage, parking lots, building management systems, etc., and it has sufficient performance and stability. Has been proven. Designed by humans around 2003OntologyDeveloped into the field of network intelligence that utilizes (expressed as fuzzy rules)[132].

2000 era

In 2005, Ray Kurzweil wrote, "Overwhelming artificial intelligence transcends humans in terms of knowledge and intelligence, advances science and technology and transforms the world.Technical singularity(Singularity) will visit in 2045."

In 2006,Jeffrey HintonBy the research teamAuto encoderbyneural networkThe deepening method ofDeep learningThe direct origin of).

Early 2010s

Since the beginning of the 2010s, an environment for research and development that handles huge amounts of data has been established, and AI-related research has begun to make great progress again.

2010 In the UK Economist magazinebig dataWas proposed. In the same yearQuestion answering system OfWatsonBut the quiz show "Jepady!It was a big news that we defeated humans in the practice game of[133].2012 In the image processing contestJeffrey HintonThe third AI boom began when his team won the victory after achieving significant accuracy improvement over the conventional method.

In 2013National Institute of Informatics[Annotation 5],Fujitsu LaboratoriesDeveloped by the research team ofToro Robot,” said that he challenged the mock examination of the University of Tokyo entrance examination. The examinees actually faced using a dedicated program for calculating mathematical expressions and analyzing wordsUniversity entrance examination center examinationAnd deciphered the question of the University of Tokyo second examination.Yoyogi Seminar"It was difficult to pass the University of Tokyo, but it was acceptable for private universities"[134].

In 2014, was a Japanese artificial intelligence scientistMotoaki SaitoHe also proposed the concept that the production cost of food, clothing, and shelter will approach zero as much as possible due to the advancement of automation and computer technology before the singularity.

Jeff HawkinsHowever, he is continuing his research toward its realization, but he is developing his own theory in his book "Thinking Brain Thinking Computer".

In various countries around the world, both military and civilian research and development are progressing toward practical use, but especially unmanned fightersUCAVAnd unmanned carsRobot carAlthough it is still under development, full automation was still experimental in the 2010s (UCAV is used, but some operations are often done from the ground).

For robots,CSAILAdvocated by Rodney BrooksInclusive architectureThe theory has appeared. This does not precede conventional knowledge of "I think, therefore I am", but uses an action-type system that learns from the environment using only the neural network of the body. Based on thisGenghisA six-legged robot, called a robot, acts as if it were alive, even though it has no so-called "brain".

Late 2010s

January 2015DeepMindCreated by the companyAlphaGoAfter defeating a professional Go playerDeep learningA method called (deep learning) is attracting attention, and in addition to research on artificial intelligence itself, research on the effects of artificial intelligence on employment etc. is also underway.[135].

In October 2016, DeepMind announces "Differentialable Neural Computer", an artificial intelligence technology that derives the relevance of input information and approximates the hypothesis.[136]Then, in November of the same year, a deep learning system that enables "one-shot learning" that does not require a large amount of data[137], The following June 2017, developed a system with human-like cognitive ability such as relational reasoning[138].. In August 2017, the symbol grounding problem (Symbol grounding problem) Was solved[139].

It is an incomplete information game that has traditionally been considered unsuitable for AI.pokerBut AI has come to beat humans[140].

An even more ambitious effort, Google officials said, is developing AI that can perform more than a million tasks with a single piece of software.[141].

The third boom of artificial intelligence: AGI (general purpose artificial intelligence) and technical singularity

The invention of deep learning in 2006 and after 2010big dataDue to the improvement of the collection environment and the high performance of the GPU, which is a computational resource, Deep Learning won the image processing contest by overwhelmingly different from other methods in 2012, and the concept of technological singularity rapidly increased. It has gained the attention of intellectuals all over the world and can now be perceived with a sense of reality. With the invention and rapid spread of deep learning, in the field of research and development,Demis HasabisLeading DeepMind, Vicarious, IBM Cortical Learning Center, whole brain architecture, PEZY Computing, OpenCog, GoodAI, nnaisense, IBM SyNAPSE, etc.General-purpose artificial intelligence(AGI) Has been launched in many projects. At these research and development sites, the brain was reverse engineered and constructed.NeuroscienceApproach that combines machine learning with machine learning is promising[142].. As a result, theories that deal with multiple tasks at the same time, such as Hierarchical Temporal Memory (HTM) theory and the updated version of Complementary Learning Systems (CLS) theory, that take one step further from deep learning that deals with only one task, are being proposed. ..

Moving a model in a virtual space such as a 3D game and learning about the real world at high speed has also achieved great results (simulationLearning by).

In addition, we think that it is impossible to reproduce intelligence with AGI alone, though there are only a few, and researchers who require whole-body simulation to reproduce body intelligence and AL (artificial life) that behaves more like living creatures There are also researchers who challenge the creation of, and researchers who challenge the digital reproduction of consciousness (artificial consciousness), which seems to be closely related to intelligence.

With the advent of large amounts of computing resources at a reasonable cost, big data has emerged, and companies are extremely interested in utilizing vast amounts of data. R&D competition on artificial intelligence is being developed by making various investments. Also in 2011D-Wave SystemsbyQuantum annealing methodWith the commercialization of the next generation, the next-generation IT infrastructure called quantum computer capable of ultra-parallel processing began to be rapidly put into practical use. The environment is starting to improve. In response to this trend, from around 2016, news programs for the general public have become prominent reports of artificial intelligence research and development, new service development, and quantum computers.

In 2017Elon MuskBut to keep humans behind the rapidly evolving artificial intelligenceBrain Machine InterfaceR & D(English editionWas announced that it had been launched.

2017 In 10 monthJeffrey HintonProposed a capsule network that enables learning including the relative positional relationship between elements.[143].

International University on March 2018, 3GLOCOMAnalysis suggests that using problem-solving AI can contribute to social change.[144].

2018 year 8 month,OpenAIAnnounces an AI that implements curiosity and performs no-game scores, no goals, no rewards, and purposeless exploration.The most human-like AI ever[145].

In September 2018, the MIT Lincoln Laboratory developed an architecture that clearly shows the steps taken to identify neural network inference, which was traditionally a black box.[146].

2019 years,BERTLanguage models such as Wikipedia have made great progress in language processing, which has been considered difficult in deep learning, and have surpassed humans in reading comprehension tests using Wikipedia and other languages.[147].

Early 2020s

2020 Is a natural language processing program with 1750 billion parameters by OpenAI(English editionWas developed and is an American bulletin board siteRedditFor a week, I continued posting and dialogue with humans without being noticed by anyone.The reason I noticed the program was not that the text was unnatural, but that the number of posts was abnormal.[148].

DeepMindDeveloped byProtein structure predictionI doAlphaFold2 CASP OfGlobal distance test Scored 90 points or more in (GDT)Computational biologyIt was an important achievement in the world, and was called a great progress toward the grand challenge of biology from decades ago.[149].

OpenAI published a paper that the performance of Transformer (language model) is a simple power of N parameters, dataset size D, and computational budget C (scaling law).

In cutting-edge AI research, a model with 2 times the size has appeared in 1000 years, and a computer with 1000 times the computing power is needed.[150].

2021 year 4 month,NVIDIA"There will be an AI model with 100 trillion parameters in the next few years," said Palais Kalya, an executive at the company.[151].

Microsoft tests 2021 trillion parameter AI in May 5[152].

In June 2021, Beijing Jiyuan Artificial Intelligence Research Institute, supported by the Chinese government, announced AI "Godo 6" with 1 trillion parameters.[153].

In June 2021, Google researchers used machine learning to create AI chips that produced floor plans for chips that were more than human-designed in all major indices such as power consumption and performance.And the time required for design was 6/1 of humans[154].

In August 2021, Hartmut Neben, who heads Google's Quantum Artificial Intelligence Research Division,Quantum computerAI was mentioned as a field that has the greatest influence on the development of machines, such as the machine learning field.[155].

2021 Currently, general-purpose artificial intelligence has not been realized,Question answering,Decision support, Demand forecast,voice recognition,Machine translation, Science and technology calculation, text summarization, etc., systems specialized in each field and frameworks that combine these have been put into practical use.[156][157][6].

Philosophy and AI


GoogleLaunched an artificial intelligence project in March 2019ethicsTo teach in the plane哲学PersonpolicyPlanner·EconomicsAnnounced that it will establish an AI ethics committee consisting of people and technologists[158].. But the ethics committeeAnti-science・AntiMinority-Skepticism about global warmingEtc. are also included, and Google employees requested dismissal[158].. On April 4, Google announced the dissolution of the ethics committee because it "turned out to be unable to function as expected"[158].

Oriental philosophyAccording to Yoichiro Miyake's theme of absorbing AI intoTakahito IguchiIs "torii(TORII)” my project, “We areanimismAnd for everythingSpiritFind a targetculture,” says Miyake and Tetsuishi Tadahiro.[159].. Animistic artificial intelligence theoryContemporary ArtOr,Zen OfEnlightenmentHow to get AI to do it?"[159].

Former Google engineer Anthony Levandowski said AI in 2017ToReligious groupFounding "Way of the Future"[160].. The mission of the organization is expressed in an abstract manner, "to promote and develop the realization of Godhead based on artificial intelligence (AI), and to contribute to the improvement of society through understanding and worship of Godhead." Many foreign mediaSFmovies,HistoryReported in association with[160].UberAnd Google's Waymo said Levandowskiself-drivingAboutConfidential informationThePlagiarismAppealing fortrialMeanwhile, Levandowski is the source of UberCEO(Travis Karanik)BotOne by one, weConquer the worldDemonstrating ambitious behavior, such as saying "I will do it."[160].

Soai UniversityProfessor of Faculty of HumanitiesShatetsu sect"The line between philosophy, thought and literature, and religion and spirituality is becoming obscure."[161].. According to Juichi Uchida, a philosopher and ethicist, "every real philosopher is talking about the dead, ghosts, and the otherworld."[162].

InventorRay KurzweilSays the philosopherJohn SarlePosed byStrong AI and weak AIControversy is a hot topic in AI philosophical debate[163].. philosopherJohn SarleandDaniel DennettAccording to Sar'sChinese room"Ned BlockRan'Chinese brain"such asFunctionalismCritical thinking experiments show that true consciousnessFormal logicClaims that the system cannot achieve[164][165].


"What does it mean to talk about science?"Yasushi SudoIs a philosophical consideration of science (Philosophy of science) Point out that it is actually "disconnected" from science[166].. Also,"Heart"Or"ConsciousnessHas been solvedBrain science-Computer science(Computer science) · Artificial intelligenceResearch and DevelopmentIn relation to etc.Francis clickCriticizes that "philosophers have had little success in the year 2000"[167]..From this point of view哲学Is evaluated as nothing more than "third-class, not second-class" scholarship and science.[167].. Brain scientistToshiyuki SawaguchiAgrees with Click and says, "This is a sigh of sighing for me, but philosophers and thinkers continueSpare time"I think."[167].. In fact, philosophy isScoray) Startedア リ ス ト テ レ スThe philosopher of science says that the negative remarks mentioned above are not irrelevant.Keiichi NoieSays[167].

Philosophers are different from science日常Have tried to talk about "existence" and "universe" in a dynamic language[168].. HoweverTheoretical physicsA personDiracDid not trust the philosopher very much[169].. Dirac saw,WittgensteinPhilosophers includingQuantum mechanicsOn the contraryPascalThe followingprobabilityDo not even understand the concept of[170].. Dirac's idea is that no matter how much non-scientific everyday language you use, you can't communicate accurately.[169].

Life informaticsPersonNeurosciencePerson ofKazuyuki AiharaAccording to the book “Artificial Intelligence is Created in this Way”Technical singularityThe ideology and philosophy of ", Singularity" are discussed in some parts, but even if it is called a singular point,MathIs not a story[171].. The above-mentioned article said, "In the first place, in the discussion related to singularity,brainOverStatementGood in itselfDefinitionIt's not done"[172].. Indeed, the brain is "digitalInformation processing systemFrom the point of view of ”, singularity may occur[173].. But the real brain is not such a simple system,デ ジ タ ルanalogWas fusedhybridTo be a "system"NeuroscienceIs shown in the observation results of[173].. According to the above article,NerveVarious types of membranesnoiseThere is, and there is an analog amount with this noise in the brainNeuronof"カ オ スIs being created, it is considered "extremely difficult" to describe this situation digitally.[174].

Mathematician-Logician OfKazuyuki Tanaka"Is a general philosopherlogic OfExpertNot"[175], Computer scientist (computer scientist) and logician Torkel ゠ Francaine, many of the mathematical references by philosophers are "terrible."Misunderstanding,Free associationIt is based on[176].. According to Tanaka,Godel's incompleteness theoremA book written by a philosopher was about to be sold in a bookstore around the same time as a book by Francaine, but the philosopher's book was badly criticized by a specialist[175].. The book was easy to read as a whole and was well received by general readers,GodelThe core of the proof of (Fixed point theorem) Was explained with a fundamental misunderstanding[175].. Similar mistakes can be found in other introductory books[175].. According to Francaine, misunderstandings and misuses about the incompleteness theorem occur in general, including philosophy.[176], Religion ortheologyBut being abused[177][178].. In 1931, Godel showed that "specificFormal systemThe existence of undecidable propositions in, and not the theorem of "imperfections" in the general sense[179].

Science and philosophy

According to "What does it mean to talk about science?", the problems dealt with by academic fields were organized and differentiated, and science and philosophy began to study different problems.[180].. This is "the subdivision of the research field itself," andProgressSays astrophysicist Sudo[180].. On the other hand, scientific philosophers and ethicistsTetsuji IsetaIs a philosophical and unified approach to dealing with "big" problems that include various elements.astronomyMentioned about[180].. "It's true, but astronomy then stopped dealing with that [philosophical] problem, and today's physics doesn't deal with it either," Sudo said, "yes, but something in itself. Is there?"[180].. Sudo also stated that[166].

"Science philosophy and the disconnection of science"

I have never heard that the philosophy of science gave any advice to physicists, and it is safe to say that the philosophy of science and general scientists are almost dead negotiations. … It is certain that there is a deep gap between the values ​​of philosophers of science and scientists.

20st centuryIs one of the greatest physicists born ofRichard Feynman"Scientific philosophy is only useful to scientists to the extent that ornithologists are useful to birds," said the famous phrase. … When I gave a lecture quoting this word, I received a counterargument that “Ornithology is not done for birds, and the philosophy of science does not exist for science either”. There is.Indeed, the philosophy of science does not have to be for science[166].
This is the philosophy of sciencemethodologyScience is always stepping forward while wondering if it's really right.Isn't that all right?Scientists say a lot from the side, but scientists just tilt their heads asking, "Is this an important point to listen to?" (Maybe the philosophers of science will hit you). It's an honest impression[181].

Sudo takes up the term "cause" that is philosophically discussed, and says, "No further discussion is possible unless the term "cause" is specifically defined."[182]"It is certain that the definition of causality that philosophers are interested in is different from that of physicists," he said.[183].. Iseta says, "The difference in the appearance of physicists and philosophers may be greater than I expected."[184].

In a dialogue, Sudo said, "I've been discussing for a long time so far. I think that the difference in opinion has been clarified, but will it really make a difference?" "I'm not going to settle," he said.[185].

Literature/Fiction/SF (Science)


[How to use footnotes]

注 釈

  1. ^ British: scruffy AI
  2. ^ British: artificial intelligence
  3. ^ British: chatterbot
  4. ^ The first successful knowledge base program in mathematics.
  5. ^ Noriko AraiIs the leader.


  1. ^ a b c Sato 2018, p. "Artificial intelligence".
  2. ^ ASCII.jp 2018, p. "Artificial intelligence".
  3. ^ Momouchi 2017, p. "Artificial intelligence".
  4. ^ a b Kodansha(2017)Artificial intelligence"A dictionary that understands IT terms", Asahi Shimbun, VOYAGE GROUP
  5. ^ Kodansha (2017) "Natural language processing"A dictionary that understands IT terms", Asahi Shimbun, VOYAGE GROUP
  6. ^ a b ""AI that summarizes any sentence in 3 lines" demo site released by a venture from the University of Tokyo Matsuo Lab "Accuracy is comparable to humans"”(Japanese). ITmedia NEWS. 2021/9/2Browse.
  7. ^ "[2th] Let's experience the hottest deep learning now (page XNUMX)”. Enterprise (January 2015, 1). 2015/5/30Browse.
  8. ^ "AI now exceeds humans in poker! What is its significance?". Gizmodo (September 2017, 2). 2018/2/7Browse.
  9. ^ "Artificial intelligence wins again in "AI vs. human" poker showdown, wins 6 days battle against 5 people and gets 3000 million yen". GIGAZINE (September 2017, 4). 2018/2/7Browse.
  10. ^ "Mahjong AI Microsoft Suphx achieves results comparable to top human players”(Japanese). News Center Japan (September 2019, 8). 2021/9/2Browse.
  11. ^ "Evolution and history of game AI”(Japanese). News Center Japan (September 2019, 8). 2021/9/2Browse.
  12. ^ Challenges to the personalization era and the impact of Creative AI
  13. ^ http://www.fiercebiotech.com/data-management/google-joins-brain-initiative-to-help-petabyte-scale-data-sets
  14. ^ http://news.mynavi.jp/articles/2016/08/10/isc2016_braininitiative/
  15. ^ http://news.mynavi.jp/articles/2017/09/11/hotchips29_google/001.html
  16. ^ “13th Five-Year Plan, China's Technology Innovation Plan Revealed”. People's network(July 2016, 7). http://j.people.com.cn/n3/2016/0728/c95952-9092181.html 2018/2/7Browse. 
  17. ^ Poo, Mu-ming; Du, Jiu-lin; Ip, Nancy Y; Xiong, Zhi-Qi; Xu, Bo and Tan, Tieniu (2016),'China Brain Project: Basic Neuroscience, Brain Diseases, and Brain-Inspired Computing ', Neuron, 92 (3), 591-96.
  18. ^ "Promote AI in All China". Nihon Keizai Shimbun(July 2017, 12). https://www.nikkei.com/article/DGKKZO24371260X01C17A2FFE000/ 2018/2/7Browse. 
  19. ^ “China selects 18 geniuses under the age of 27 and puts them into the development of AI weapons”. Newsweek(July 2018, 11). https://www.newsweekjapan.jp/stories/world/2018/11/ai-41.php 2018/11/29Browse. 
  20. ^ a b “Chinese one-party dictatorship, advantageous for AI development competition”. Wall Street Journal(July 2018, 2). http://jp.wsj.com/articles/SB12346302927663484593304584068193706085748 2018/3/7Browse. 
  21. ^ “Chinese chasing in AI development race, threatening US company lead”The Wall Street Journal. (January 2018, 1). http://jp.wsj.com/articles/SB10806998528272603825204583646830641781270 2018/2/7Browse. 
  22. ^ "China's "Super AI surveillance society"---in Xinjiang Uygur Autonomous Region, monitor even the "body"! ”. Shueisha(July 2018, 2). https://wpb.shueisha.co.jp/news/technology/2018/02/03/99109/ 2018/2/7Browse. 
  23. ^ "Computer hegemony seen from the first place of Green500". Ube Industries. 2018/3/7Browse.
  24. ^ “Testing face recognition system operation in Xinjiang, China. Warns authorities if more than 300m away from designated area”. Engadget(July 2018, 1). http://japanese.engadget.com/2018/01/19/300m/ 2018/2/7Browse. 
  25. ^ "No one can stop the move of China becoming an "AI superpower.". WIRED(July 2017, 8). https://wired.jp/2017/08/16/america-china-ai-ascension/ 2018/2/7Browse. 
  26. ^ “World Face Recognition Benchmark Test Results Announced China Top 5 Monopoly”. Chinese net(July 2018, 11). http://japanese.china.org.cn/culture/2018-11/24/content_74198278_0.htm 2018/12/3Browse. 
  27. ^ “Chinese team monopolizes top ranks in AI's world champion deciding match “ImageNet””Forbes. (August 2017, 8). https://forbesjapan.com/articles/detail/17213 2018/2/11Browse. 
  28. ^ "THE AI COLD WAR THAT COULD DOOM US ALL". Forbes(July 2017, 8). https://www.wired.com/story/ai-cold-war-china-could-doom-us-all/ 2018/2/7Browse. 
  29. ^ Politico Magazine Staff (September 2018, 9). “The POLITICO 50 Reading List”. Political. https://www.politico.com/interactives/2018/politico-50-book-picks/ 2018/10/9Browse. 
  30. ^ "French President Macron declares "AI nation", unmanned autonomous driving also lifted" (Japanese). Forbes JAPAN(July 2018, 3). https://forbesjapan.com/articles/detail/20401 2018/4/5Browse. 
  31. ^ https://www.sbbit.jp/article/cont1/35264?page=2
  32. ^ "The number of papers on artificial intelligence, the top three in the US and China"Nihon Keizai Shimbun. (June 2017, 11). https://www.nikkei.com/article/DGXMZO22943380R31C17A0TJU200/ 2018/2/7Browse. 
  33. ^ "Speeding up the creation of "musical music" for music games with AI KLab uses "Skusta" to halve the required time”(Japanese). ITmedia NEWS. 2021/9/1Browse.
  34. ^ "NVIDIA Canvas: Leverage the power of AI”(Japanese). NVIDIA. 2021/7/10Browse.
  35. ^ "AI that "draws anime pictures automatically" appears-will the work of animators be taken away?”(Japanese). ITmedia Business Online. 2021/7/10Browse.
  36. ^ "Colorization of manga, AI is useful even if the accuracy is not 100%”(Japanese). ITmedia NEWS. 2021/11/8Browse.
  37. ^ a b Watanabe double crown is also adopted in the war of the king!What is the recently popular method of building an elmo enclosure? [How to enclose balls 63rd] --Japan Shogi Association delivered on March 2019, 3
  38. ^ a b c https://www.itmedia.co.jp/news/articles/2010/29/news084_3.html
  39. ^ Geraci 2012, pp. 38–39.
  40. ^ Geraci 2012, pp. 1–2.
  41. ^ a b c Geraci 2012, p. 64.
  42. ^ Dr. Hawking "The evolution of artificial intelligence means the end of humankind"
  43. ^ "It's like calling the devil," Elon Musk talks about the dangers of artificial intelligence
  44. ^ Bill Gates also worried about threat of artificial intelligence
  45. ^ ""Father of Deep Learning" Calls for AI Use in China". Sankei Biz(July 2019, 4). https://www.sankeibiz.jp/macro/news/190401/mcb1904010710001-n1.htm 2019/4/5Browse. 
  46. ^ “Deep Learning'Godfather' Bengio Worries About China's Use of AI”. Bloomberg(July 2019, 2). https://www.bloomberg.com/news/articles/2019-02-02/deep-learning-godfather-bengio-worries-about-china-s-use-of-ai 2019/4/5Browse. 
  47. ^ a b "Freedom on the Net 2018 The Rise of Digital Authoritarianism". Freedom House. 2018/12/13Browse.
  48. ^ "Artificial intelligence and digital authoritarianism: can democracy survive?". Foreign Affairs. 2019/11/10Browse.
  49. ^ "China's digital dictatorship". economist (September 2018, 7). 2019/11/10Browse.
  50. ^ "China's digital dictatorship, ending ends beyond novel". Wall Street Journal (September 2018, 7). 2019/11/10Browse.
  51. ^ "China's NPC Opening Adjacent to digital dictatorship". Mainichi Newspapers (September 2018, 7). 2019/11/10Browse.
  52. ^ "Does China's digital police state have echoes in the West?". economist (September 2018, 6). 2019/11/10Browse.
  53. ^ "The rise of China as a digital totalitarian state". Washington Post (September 2018, 2). 2019/11/10Browse.
  54. ^ "AI dictatorship scattered in China Surveillance system expanded in sales". Nihon Keizai Shimbun (September 2019, 6). 2019/11/12Browse.
  55. ^ "The Chinese government is promoting a plan to directly retrieve data from workers' brains (China)". Carapaya (September 2018, 5). 2019/1/11Browse.
  56. ^ "A'emotional monitoring system' that scans and manages workers' brain waves has been developed in China and actually put into the field". GIGAZINE (September 2018, 5). 2018/5/7Browse.
  57. ^ "Chinese companies monitor Emotions of employees with EEG helmets". MIT Technology Review (September 2018, 5). 2018/5/1Browse.
  58. ^ "'Forget the Facebook leak':China is mining data directly from workers' brains on an industrial scale". South China Morning Post (September 2018, 4). 2018/5/11Browse.
  59. ^ "[FTAI is increasing, in China's Internet censorship work] ”. Nihon Keizai Shimbun (September 2019, 11). 2018/5/23Browse.
  60. ^ "Angle: 30 years since the Tiananmen Incident, China is wary, "AI censorship" full operation". Reuters (September 2019, 5). 2019/11/12Browse.
  61. ^ “Is China's corruption-busting AI system'Zero Trust' being turned off for being too efficient?”. South China Morning Post(July 2019, 2). https://www.scmp.com/news/china/science/article/2184857/chinas-corruption-busting-ai-system-zero-trust-being-turned-being 2019/11/12Browse. 
  62. ^ What is the result of introducing "an AI system that detects corruption of bureaucrats" in China? ”. GIGAZINE(July 2019, 2). https://gigazine.net/news/20190204-china-corruption-busting-ai-system/ 2019/11/12Browse. 
  63. ^ "No escape? Chinese VIP jail puts AI monitors in every cell'to make prison breaks impossible'". South China Morning Post (September 2019, 4). 2019/4/11Browse.
  64. ^ “Chinese Social Score AI Determines “Ignore Signals” for Bus Advertising Women”. Gizmodo(July 2018, 11). https://www.gizmodo.jp/2018/11/facial-recognition-flags-woman-on-bus-ad.html 2019/11/12Browse. 
  65. ^ “Jaywalkers under surveillance in Shenzhen soon to be punished via text messages”. South China Morning Post(July 2018, 3). https://www.scmp.com/tech/china-tech/article/2138960/jaywalkers-under-surveillance-shenzhen-soon-be-punished-text 2019/11/12Browse. 
  66. ^ "Pedestrian signal disregard detected by face recognition, reflected in credit record Nanjing". AFPBB(July 2019, 7). https://www.afpbb.com/articles/-/3234320 2019/11/12Browse. 
  67. ^ "Chinese police are strengthening the surveillance system by introducing a sunglasses type device equipped with a face recognition function”. GIGAZINE (May 2018, 2). 2019/1/11Browse.
  68. ^ "Robots run the city of the future? Reality in Dubai and China". with the BBC (September 2018, 6). 2019/4/25Browse.
  69. ^ "Following safety specific name - 習近 Heijikuni "黑科 skills" Yes somewhat?". Voice of America (September 2019, 11). 2019/11/11Browse.
  70. ^ "Focus: China builds surveillance nation using black technology". Reuters (September 2018, 3). 2019/11/11Browse.
  71. ^ "China strengthens high-tech surveillance Uses AI to control public opinion". Nihon Keizai Shimbun (September 2019, 11). 2019/11/10Browse.
  72. ^ "China strengthens surveillance with AI". Nihon Keizai Shimbun (September 2019, 5). 2019/11/10Browse.
  73. ^ “Exposed: China's Operating Manuals For Mass Internment And Arrest By Algorithm”. International Research Journalists Union(ICIJ)(July 2019, 11). https://www.icij.org/investigations/china-cables/exposed-chinas-operating-manuals-for-mass-internment-and-arrest-by-algorithm/ 2019/11/26Browse. 
  74. ^ “Large-scale system monitors Uighurs reveals internal documents of Chinese authorities”. Tokyo Newspaper(July 2019, 11). https://www.tokyo-np.co.jp/article/world/list/201911/CK2019112502000118.html 2019/11/26Browse. 
  75. ^ “'Automated Racism': Chinese Police Are Reportedly Using AI to Identify Minority Faces”. Gizmodo(July 2019, 4). https://gizmodo.com/automated-racism-chinese-police-are-reportedly-using-a-1834054068 2019/4/16Browse. 
  76. ^ “One Month, 500,000 Face Scans: How China Is Using AI to Profile a Minority”. New York Times(July 2019, 4). https://www.nytimes.com/2019/04/14/technology/china-surveillance-artificial-intelligence-racial-profiling.html 2019/12/10Browse. 
  77. ^ “Secret documents: China detention camps to'prevent escapes'”. Mainichi Newspapers(July 2019, 11). https://mainichi.jp/english/articles/20191125/p2g/00m/0in/039000c 2019/11/25Browse. 
  78. ^ “Hong Kong citizens afraid of surveillance society fear mainland China-police uses face recognition AI”. Bloomberg(July 2019, 10). https://www.bloomberg.co.jp/news/articles/2019-10-24/PZTF9HT0AFBB01 2019/11/11Browse. 
  79. ^ "Chinese-style surveillance vs. young people's wisdom Hong Kong demonstrations, even in the cyber realm". Nihon Keizai Shimbun (September 2019, 6). 2019/6/5Browse.
  80. ^ "Hong Kong demonstration, participants' new target is "smart streetlight" with surveillance cameras". Business insider (September 2019, 8). 2019/11/12Browse.
  81. ^ "China exports AI surveillance technology to 54 countries worldwide". Newsweek (September 2019, 4). 2019/4/26Browse.
  82. ^ "China Is Taking Its AI Around The World. This Should Scare The US”. Medium (August 2018, 8). 2018/10/26Browse.
  83. ^ "“Face recognition AI” that is being put to practical use in China is spreading throughout the world". Forbes (September 2017, 12). 2018/10/26Browse.
  84. ^ "China exports its high-tech authoritarianism to Venezuela.It must be stopped.". Washington Post (September 2018, 12). 2018/12/13Browse.
  85. ^ "How China's AI Technology Exports Are Seeding Surveillance Societies Globally”. THhe Diplomat (October 2018, 10). 2018/10/26Browse.
  86. ^ "Chinese tech groups shaping UN facial recognition standards". Financial times (September 2019, 12). 2019/12/10Browse.
  87. ^ "China spurs on “AI standardization” and proposes international standards". Forbes (September 2019, 12). 2019/12/10Browse.
  88. ^ "AI judges your credit rating, rating 14 billion people in Japan and China". Asahi Shimbun Gloves. 2019/12/1Browse.
  89. ^ "Fear of "downstream" AI judgment accelerated in China". President Online(July 2016, 7). https://president.jp/articles/-/24771 2019/2/22Browse. 
  90. ^ "Amazon and MIT Researchers Conflict over "Difference in Amazon Face Recognition Software"". GIGAZINE(July 2019, 4). https://gigazine.net/news/20190415-amazon-rekognition-mit-researcher/ 2019/4/16Browse. 
  91. ^ "[Breaking news] Boston Dynamics announces new functions and application examples of quadruped robot "Spot" Autonomous movement / remote control / operation management / strengthening arm | Robosta-Robot Information WEB Magazine”(Japanese). Robosta. 2021/7/10Browse.
  92. ^ a b https://wired.jp/2021/05/02/new-york-returns-police-robodog-after-public-outcry/
  93. ^ "Lethal Robotic Technologies: The Implications for Human Rights and International Humanitarian Law" Philip Alston, Journal of Law, Information and Science, 2012
  94. ^ “Israeli's “auto-kill zone” spreads across the Gaza Strip”. WIRED(July 2008, 12). https://wired.jp/2008/12/08/ガザ地区境界線に広がる、イスラエルの「自動殺傷ゾーン」/ 2019/4/16Browse. 
  95. ^ a b Artificial Intelligence and Chinese Power(Foreign Affairs August 2017, 12)
  96. ^ "AI pilot with Raspberry Pi, win human pilot in simulated fight using military flight simulator”. Business newsline. 2016/9/19Browse.
  97. ^ ""Top Gun" lost to AI, but unilaterally shot down in a pseudo-air war "double hard for PC for children"”. Sankei WEST. 2016/9/19Browse.
  98. ^ [1]
  99. ^ cnn.co.jp-Warning letter for military use of artificial intelligence, open letter by celebrities such as E. Musk 2015.07.30 Thu posted at 11:51 JST
  100. ^ “Continuing debate over murder robot weapons, regulation UN ends first expert meeting”. Sankei News(July 2017, 11). https://www.sankei.com/world/news/171118/wor1711180027-n1.html 2019/4/16Browse. 
  101. ^ "Concern over "AI weapons" international rule agreement is not legally binding". NHK(July 2019, 8). https://www3.nhk.or.jp/news/html/20190822/k10012044131000.html 2019/8/22Browse. 
  102. ^ ""Military use of artificial intelligence" to replace nuclear weapons, the current state of development races in the three countries”. WIRED (June 2017, 9). 2018/2/7Browse.
  103. ^ "AI weapons development, the US and China spark spark North American attack CG". Asahi Shimbun (September 2018, 12). 2019/2/22Browse.
  104. ^ "For attack and reconnaissance... Chinese army develops unmanned boats with XNUMX ships, "herd" World's largest experiment". ZAKZAK (September 2018, 6). 2019/3/5Browse.
  105. ^ "U.S. military uniforms top fears Google". Nihon Keizai Shimbun (September 2019, 3). 2019/4/16Browse.
  106. ^ "Google is helping China, not the United States,-Trump criticizes". Nihon Keizai Shimbun (September 2019, 3). 2019/4/16Browse.
  107. ^ "(The Economist) Business Justice of US Companies". Nihon Keizai Shimbun (September 2019, 4). 2019/4/16Browse.
  108. ^ "A fresh internal e-mail leaked by a person involved in a confidential plan ``Project Maven'' that uses Google's AI militaryly". gigazine (September 2018, 6). 2019/4/16Browse.
  109. ^ "2018 tech world lie worst 8". Gizmodo (September 2018, 12). 2019/4/16Browse.
  110. ^ "Google denies link to China's military over touch-screen tools that may help PLA pilots”. (June 2019, 7). 2019/7/5Browse.
  111. ^ "[FTMicrosoft collaborates with Chinese military on AI research] ”. Nihon Keizai Shimbun (September 2019, 4). 2019/4/11Browse.
  112. ^ "SecDef: China Is Exporting Killer Robots to the Mideast". Defense One (September 2019, 11). 2019/11/5Browse.
  113. ^ ""AI hacker" threat, China already commercialized?". Sankei News (September 2019, 1). 2019/1/17Browse.
  114. ^ "Artificial intelligence is at a stage of abuse”. BBC News (October 2018, 2). 2018/3/3Browse.
  115. ^ Masayuki Yuda (May 2016, 5). “Introducing Robo to raise wages? McDonald's “Battle of $15 per hour”” (Japanese). Nihon Keizai Shimbun. https://www.nikkei.com/article/DGXLASDZ27H3C_X20C16A5000000/ 2019/9/30Browse. 
  116. ^ Michael A. Osborne; Carl Benedikt Frey (2013-09-13). THE FUTURE OF EMPLOYMENT: HOW SUSCEPTIBLE ARE JOBS TO COMPUTERISATION?. Oxford Martin Program on Technology and Employment. Pp. 70. https://www.oxfordmartin.ox.ac.uk/publications/the-future-of-employment/ 2019/10/6Browse.. 
  117. ^ Megumi Ueda (2017-12-05) (PDF). A future that coexists with AI-human resources in the age of AI- (Report). Co., Ltd.Nomura Research Institute. https://www.mhlw.go.jp/file/05-Shingikai-12602000-Seisakutoukatsukan-Sanjikanshitsu_Roudouseisakutantou/0000186905.pdf 2020/8/30Browse.. 
  118. ^ Ministry of Internal Affairs and Communications (2018-07). 2018 White Paper on Information and Communication Part 1 Special Feature Sustainable Growth by ICT in the Age of Declining Population Section 5 Future Work Due to Evolution of ICT (1) Occupational Changes (Report). https://www.soumu.go.jp/johotsusintokei/whitepaper/ja/h30/html/nd145210.html 2020/6/5Browse.. 
  119. ^ Koichi Iwamoto; Yuta Tagami (2018-05). “Impact of artificial intelligence AI on employment; actual conditions in Japan” (Japanese). Policy Discussion Paper (Japanese) (Research Institute of Economy, Trade and Industry) P-18-009: 1-17. https://www.rieti.go.jp/jp/publications/summary/18050004.html 2020/6/5Browse.. 
  120. ^ Melanie Arntzi; Terry Gregoryi; Ulrich Zierahni (2016-5-14). “The Risk of Automation for Jobs in OECD Countries” (English). OECD Social, Employment and Migration Working Papers (OECD) 189: 1-34. two:10.1787 / 1815199X. ISSN 1815199X. https://www.oecd-ilibrary.org/social-issues-migration-health/the-risk-of-automation-for-jobs-in-oecd-countries_5jlz9h56dvq7-en 2020/6/5Browse.. 
  121. ^ "Human VS Computer Othello 6 years after the total loss of 20 shocks, I asked former world champion Ken Murakami "What I saw after losing"". ITmedia (September 2017, 10). 2018/12/25Browse.
  122. ^ a b Motohide, Mano, Isao, Hayashi "Current status and prospects of fuzzy neural networks (<Special feature> Fuzzy neural networks)"Journal of the Japan Fuzzy Society," Vol. 5, No. 2, April 1993, 4, 15–178, two:10.3156 / jfuzzy.5.2_178, ISSN 0915-647X.
  123. ^ "Sendai Subway Railway Total System”. Hitachi. 2020/11/23Browse.
  124. ^ a b "Fuzzy home appliances and neuro-fuzzy home appliances born from Matsushita Electric”. Kansai University. 2019/5/12Browse.
  125. ^ "Fuzzy fully automatic washing machine (Matsushita Electric Industrial Co., Ltd.) |". www.j-soft.org. 2019/5/2Browse.
  126. ^ "Neuro-fuzzy vacuum cleaner (Hitachi, Ltd.) | Japan Society for Intelligent Information Fuzzy". www.j-soft.org. 2019/5/5Browse.
  127. ^ "Gas Room Air Conditioner SN-A4541U-D SN-A4541RF-D Operation Manual”. Tokyo Gas. 2019/5/3Browse.
  128. ^ Mitsuyuki, Kiuchi, Shinji, Kondo "Introduction of fully automatic washing machine "Aizuma-go Day Fuzzy" (NA-F 50 Y 5)"Journal of the Japan Fuzzy Society," Vol. 2, No. 3, April 1990, 8, 15–384, two:10.3156 / jfuzzy.2.3_384, ISSN 0915-647X.
  129. ^ Kaoru, Hirota "Fuzzy home appliances, where is fuzzy"The Institute of Electrical Engineers of Japan, Vol. 111, No. 5, May 1991, 5, 20–417, two:10.11526 / ieejjournal1888.111.417, ISSN 0020-2878 .
  130. ^ Sadao, Akishita "Special Feature "Application of Neuro and Fuzzy to Robots""Journal of the Robotics Society of Japan, Vol. 9, No. 2, April 1991, 4, 15–203, two:10.7210 / jrsj.9.203, ISSN 0289-1824 .
  131. ^ "By the way, there was a word like fuzzy”(Japanese). 2019/5/2Browse.
  132. ^ Toru, Yamaguchi "Neuro-fuzzy control and network intelligence"Measurement and Control," Vol. 42, No. 4, April 2003, 4, 10–321, two:10.11499 / sicejl1962.42.321, ISSN 0453-4662 .
  133. ^ Artificial intelligence challenges quiz king! Sequel finally the decisive battle-NHK online
  134. ^ Artificial intelligence challenges the University of Tokyo model trial "level of passing private university": Nihon Keizai Shimbun, browsing July 2017, 7
  135. ^ 28 White Paper on Information and Communication Chapter 4, Sections 2-4 "28 White Paper on Information and Communication (PDF version)". Ministry of Internal Affairs and Communications. 2016/9/6Browse.
  136. ^ http://ascii.jp/elem/000/001/249/1249977/
  137. ^ https://www.technologyreview.jp/s/12759/machines-can-now-recognize-something-after-seeing-it-once/
  138. ^ http://gigazine.net/news/20170616-deepmind-general-ai/
  139. ^ https://www.nikkan.co.jp/articles/view/00439317
  140. ^ "AI finally beats professional poker players in 6-player poker”(Japanese). GIGAZINE. 2021/2/22Browse.
  141. ^ Active, Nikkei Cross Tech. “"Universal AI" that Google aims at, an amazing development method of a model that can play a million roles”(Japanese). Nikkei Cross Tech Active. 2021/2/22Browse.
  142. ^ http://wba-initiative.org/1653/
  143. ^ Hinton, Geoffrey E .; Frosst, Nicholas; Sabour, Sara (2017-10-26). Dynamic Routing Between Capsules. https://arxiv.org/abs/1710.09829v2. 
  144. ^ ""Problem-solving" AI will change Japanese society: GLOCOM, an international university, announces the results of a survey of AI utilization”. @IT (March 2018, 3). 2018/3/24Browse.
  145. ^ https://thenextweb.com/artificial-intelligence/2018/08/23/researchers-gave-ai-curiosity-and-it-played-video-games-all-day/
  146. ^ https://news.mit.edu/2018/mit-lincoln-laboratory-ai-system-solves-problems-through-human-reasoning-0911
  147. ^ xTECH (Cross Tech), Nikkei. “AI finally exceeds humans even in text reading, rapid growth in one year from Google's announcement of "BERT"”(Japanese). Nikkei xTECH. 2019/11/15Browse.
  148. ^ https://gigazine.net/news/20201008-gpt-3-reddit/
  149. ^ Robert F. Service, 'The game has changed.' AI triumphs at solving protein structures, Science, 30 November 2020
  150. ^ "Cerebras Announces AI Supercomputer for Learning GPT-3 in One Day-Hot Chips 1”(Japanese). TECH + (September 2021, 9). 2021/11/8Browse.
  151. ^ "NVIDIA joins CPU entry arm and AI calculation 10 times faster”(Japanese). Nihon Keizai Shimbun (September 2021, 4). 2021/4/13Browse.
  152. ^ "Training an AI model with 32 trillion parameters at Microsoft's ZeRO-Infinity Library”(Japanese). InfoQ. 2021/11/8Browse.
  153. ^ "Chinese research team announces new AI "Godo 2.0", number of parameters is 1 trillion, surpassing Google and OpenAI models”(Japanese). GIGAZINE. 2021/11/8Browse.
  154. ^ "Computer Science: Artificial Intelligence Speeds up Computer Chip Design | Nature | Nature Portfolio”(Japanese). www.natureasia.com. 2021/7/10Browse.
  155. ^ "Google executives, AI beyond human knowledge "advance" with quantum computer”(Japanese). Nihon Keizai Shimbun (September 2021, 8). 2021/11/8Browse.
  156. ^ Impress Corporation (September 2021, 4). “NVIDIA Jensen Huang CEO, interactive AI service "Jarvis" demo looking for "Jangara Ramen"”(Japanese). Car Watch. 2021/4/15Browse.
  157. ^ Impress Corporation (September 2021, 4). “Adobe Premiere Pro, transcription from voice is evolving.Template expansion”(Japanese). AV Watch. 2021/4/29Browse.
  158. ^ a b c Will Knight 2019.
  159. ^ a b Takahashi Mirei 2019, p. Part XNUMX.
  160. ^ a b c Tsukamoto Navy 2017, p. October 2017, 10 5:20.
  161. ^ Uchida・Shaketsu Sect (2013) "Modern Spirituality" (Kodansha Library), Kodansha, page 178
  162. ^ Uchida, J. and Shatoretsu (2013) "Modern Spirituality" (Kodansha Bunko), Kodansha, 155 pages
  163. ^ Kurzweil, Singularity (2005) p. 260
  164. ^ Searle, John (1980), "Minds, Brains and Programs", Behavioral and Brain Sciences, 3 (3): 417–457, doi:10.1017/S0140525X00005756,
  165. ^ Daniel Dennett (1991). "Chapter 14. Consciousness Imagined". Consciousness Explained. Back Bay Books. pp. 431–455.
  166. ^ a b c Yasushi SudoTetsuji Iseta(2013) "What does it mean to talk about science?---Suggest things to scientists and philosophers"Kawade Shobo Shinsha, Pages 14-15.
  167. ^ a b c d Noya 2002, p. 34.
  168. ^ All 2014, p. 178.
  169. ^ a b All 2014, pp. 177–178.
  170. ^ All 2014, p. 177.
  171. ^ Aihara et al. 2017, p. 34.
  172. ^ Aihara et al. 2017, p. 38.
  173. ^ a b Aihara et al. 2017, p. 42.
  174. ^ Aihara et al. 2017, pp. 46–47.
  175. ^ a b c d France 2011, p. 233.
  176. ^ a b France 2011, p. 4.
  177. ^ France 2011, p. 7.
  178. ^ France 2011, p. 126.
  179. ^ France 2011, p. 230.
  180. ^ a b c d Yasushi Sudo, Tetsuji Iseta (2013) “What does it mean to talk about science?--Speak to scientists and philosophers,” Kawade Shobo Shinsha, pp. 57-58.
  181. ^ Yasushi Sudo, Tetsuji Iseta (2013) "What does it mean to talk about science --- Apply to scientists and philosophers," Kawade Shobo Shinsha, 260 pages.
  182. ^ Yasushi Sudo, Tetsuji Iseta (2013) "What does it mean to talk about science?--Speak to scientists and philosophers," Kawade Shobo Shinsha, pages 124-125.
  183. ^ Yasushi Sudo, Tetsuji Iseta (2013) "What does it mean to talk about science --- Apply to scientists and philosophers," Kawade Shobo Shinsha, 141 pages.
  184. ^ Yasushi Sudo, Tetsuji Iseta (2013) "What does it mean to talk about science --- Apply to scientists and philosophers," Kawade Shobo Shinsha, 262 pages.
  185. ^ Yasushi Sudo, Tetsuji Iseta (2013) "What does it mean to talk about science --- Apply to scientists and philosophers," Kawade Shobo Shinsha, 284 pages.


Academic books and dictionaries

  • "Artificial intelligence"The Complete Encyclopedia of Japan (Nipponika)"Shogakukan-Asahi Shimbun-VOYAGE GROUP, 2018.2018/8/16Browse. “Artificial intelligence is a field of computer science that studies “intelligence” using the concept of “computation” and the tool of “computer”. To put it simply, without fear of misunderstanding, it is possible to say, "What kind of procedure (algorithm) and what kind of data (preliminary If we prepare information and knowledge, we can do it mechanically”. "
  • Sengoku, Masakazu"The importance of continuing basic research (100 years supported by the 100th anniversary special feature" Basics and Boundaries ", the next 100 years-words to give to you who will bear the future 100 years)" "Journal of the Institute of Electronics, Information and Communication Engineers (The journal) of the Institute of Electronics, Information and Communication Engineers), Vol. 100, No. 6, Institute of Electronics, Information and Communication Engineers, 2017, pp. 431-439.
  • All, Takuki "Exotic Quantum: A Strange but Unexpected Talk of Quantum", University of Tokyo Press, 2014.ISBN 978-4130636070 .
  • Noe, Keiichi"Philosophy in the Age of Science: Is Philosophy" Second-class Science "? The Quest for Philosophy, Vol. 29, Forum of Young Philosophy Researchers, 2002, pp. 31-42.
  • Francaine, Torquel, "Gedel's Theorem: An Incomplete Guide to Use and Misuse," Translated by Kazuyuki Tanaka, Misuzu Shobo, 2011.ISBN 978-4622075691 .
  • Momouchi, Yoshio"Artificial intelligence"The Complete Encyclopedia of Japan (Nipponika)", Shogakukan, Asahi Shimbun, VOYAGE GROUP, 2017.2017/12/31Browse. "Research field on the design and implementation of intelligent information processing systems using computers"
  • ASCII.jp"Artificial intelligence"ASCII.jp Digital Term Dictionary" ASCII.jp, Asahi Shimbun, VOYAGE GROUP, 2018.2018/8/16Browse. "A technology that causes a computer to perform intellectual actions such as language understanding, reasoning, and problem solving on behalf of humans."
  • Geraci, Robert M. (2012). Apocalyptic AI: Visions of Heaven in Robotics, Artificial Intelligence, and Virtual Reality (Reprinted ed.). Oxford University Press. ISBN 9780199964000 

News report

Related item

Education, research and development

R & D / Applied Science

Development examples / application examples

Research subject

Related fields

Philosophical items about AI

外部 リンク

Machine learning

Machine learning(Kikaigakushu,British: machine learning) Is a computer algorithm or its research area that automatically improves by learning from experience.[1][2],Artificial intelligenceIs considered to be a type of. Learning is done using data called "training data" or "learning data", and some tasks are performed using the learning results.For example, in the pastSpam mailIt is possible to learn by using as training data and perform the task of spam filtering.

Machine learning is closely related to the following areas:

The name machine learning was named in 1959Arthur SamuelCoined by[5].



The following concise definition by Tom M. Mitchell is widely cited, although the definition varies from one author to another:

A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P, if its performance at tasks in T, as measured by P, improves with experience E[6].
Computer program from experience E with respect to task class T and performance indicator PLearningThen, it means that the performance measured by P of the task in T is improved by experience E. — (English edition

heretaskIs a problem to be solved by the program. For example, in the case of a sales forecasting task, it is a task such as "forecast tomorrow's sales".

経 験Is given to the program as some data.This dataTraining dataOrTraining dataIn the case of a sales forecasting task, for example, sales up to today, which is "past experience", are given as training data.The process of improving the performance of a program using training data, "ProgramTraining"Do" or "ProgramLearningLet me do it. "Also, a set of all data used to train a program (training or learning)data set(Data setTomo).

FinallyPerformanceIs an index for measuring how much performance the program has achieved the task, and in the case of the above-mentioned sales forecasting task, for example, an error from the actual sales can be used as the performance index.

Variable type

In machine learning, dataxWhen is a continuous quantityxTheQuantitative variables(quantitative variable), A variable that represents the type of thing, such as a classification category such as "dog" or "cat".Qualitative variables(qualitative variable)[7][8]..Qualitative variablesCategorical variable(categorical variable),factor(factorAlso called)[8].

In addition to quantitative and qualitative variables, it takes discrete values ​​ordered as "large", "medium", and "small".Ordered categorical variable(ordered categorical variable)[8]. AlsoNatural languageUnlike qualitative variables, machine learning handles things that are not continuous quantities and do not take values ​​in a finite number of categories, unlike categorical variables.

Types of machine learning tasks

Machine learning tasks can be divided into the following three typical categories.However, these three do not cover all the tasks handled by machine learning, and some tasks belong to multiple categories and some tasks are ambiguous as to which category they belong to.

Supervised learning
Inputs and corresponding outputs[Note 2] Generate a function that maps.For exampleSortThe problem is given an example shown in the classification corresponding to the input vector and the output, and the function that maps them is approximated.
Unsupervised learning
Build a model from input only (unlabeled example).Data MiningSee also
Reinforcement learning
Learn how to act by observing the surrounding environment. Actions always affect the environment, and feedback from the environment in the form of rewards is used as a guide for learning algorithms. For exampleQ learningThere is.

Supervised learning


Supervised learning(supervised learning), Unknown probability distributionTarget.In some sense in practical applicationxTheinput,yTheoutputIn many cases, for exampley ThexUnknown functionFValueF(x)There is a small noise on it.The algorithm isFollowxyPair ofIs given as training data.The task that the algorithm should solve does not belong to (maybe) training dataxOn the other hand, the conditional probability distributionOr a value determined from it (for exampleIs to approximate the expected value of[9]..The accuracy of the approximation is predeterminedLoss functionEvaluate using the function.Therefore, it can be said that the goal of supervised machine learning is to reduce the expected value of the loss function.

I mentioned earlierDefinition of machine learningAccording to, supervised machine learning can be said to be the following machine learning:

task経 験Performance
Or approximate the value determined from it wellTraining dataExpected value of loss function

Prior knowledge in supervised learningFrom the unknownxCorresponding toyDistributionIs required to guess.Therefore, the algorithm is unknownxからThe operation to find (or the value determined from it)GeneralizationOrinference(inference).Depending on the task, it may be called "prediction", "judgment", "recognition", etc.

Algorithm is unknown dataxからxCorresponding toyIt is necessary to infer information on the distribution of, but the training data given as prior knowledge for this inferencexiMust be inferred fromyiIs attached as the "answer". The name "supervised learning" is thus a known "problem"xi"Answer" toyiThe algorithm that is a "student" is unknown in the setting that the "teacher" teachesx"Answer" corresponding toyIt is named after inferring.For the same reason, in supervised learning, training data is used.Teacher dataAlso called.

Training phase and generalization phase

In many supervised machine learning models, before the actual generalizationTrainingOrLearningA work called "training algorithm" and "generalization algorithm" can be regarded as a pair of machine learning models.The training algorithm takes the training data as input andThe parameterValue calledθIs output.The parameter is intuitively the "learning result" obtained by extracting useful information from the training data, and is this "learning result" in the case of generalization.θGeneralize using.In other words, the generalization algorithm is inputxBesides parametersθAlso it receives as input,Find (or a value determined from it).

Variable name

Variables in supervised machine learningxTheExplanatory variable(explanation variable),yTheObjective variable,Target variable(target variable) Ortarget(target)[7]..These are often referred to by different names,xThePredictive variables(predictor),yTheResponse variable(response variable)[8],xTheIndependent variable(Independent variable),yTheDependent variable(dependent variable) May be called[8]..Also, depending on the task, it may be called by a name other than these.

Regression and classification

Regression and classification are typical tasks that belong to supervised learning.Objective variable in supervised learningyWhen is a quantitative variableRegression(regression), If it is a categorical variable that takes a value in a finite setSort(classification) OrDiscriminationCall[8][10].


Regression goal is inputxWhen givenExpect information about.Typically

likeyIs an unknown functionFStatueF(x)Random noiseεInput in the case of data withxからyAs accurate as possible forecastIs required to be output.Objective variables handled in regressionyIs a continuous quantity, typically a numerical vector in which a plurality of real numbers are arranged.

Like other supervised machine learning algorithms, regression algorithmsA set of training data selected according toCan be received as, and input with these training data as hintsxCorresponding toyExpected value of

Is output.Forecast accuracy is a loss functionMeasured by.Loss function in regressionasSquare error loss

Is often used.

The goal of regression isGeneralization error(Prediction error,Forecast lossBoth)

Is to keep it small.hereIs the output of the generalization algorithm,E[・]Represents the expected value.


In the classification task, a finite number of predetermined classes are defined, and each class has "cat", "dog", etc.Class label(Or simplylabel) Is assigned a class name.The purpose of the classification task is given inputxIt is to guess which one belongs to.

There are roughly two types of algorithms for solving classification tasks: a "deterministic approach" and a "stochastic approach".[11], The former is input in the classification taskxWhen givenxIt outputs the class label to which it seems to belong, and is typically a loss function.0-1 loss


The latter, on the other hand, does not output the class label directly,Confidence(confidence scoreIs to be output.here Thex jA measure of how confident you are in the second class,Meet

Training data for classification tasks that output confidence OfyiIs also encoded so that it is consistent with certainty.That is,xi jIf it belongs to the second classAnd.hereej ThejA vector in which the third component is 1 and the other components are 0 (thus a vector in which only one component is 1 and the others are 1).one-hot vectorTo express data by one-hot vectorone-hot expression[13] ).Typically as a loss functionCross entropy


Relationship between regression and classification

A typical method for designing an algorithm for a classification task using conviction is to use the algorithm for a regression task.That is, training data in which the class is encoded by a one-hot vector.It is a method of training the algorithm of the regression task using and using the algorithm of the training result for the classification task.However, regression task outputIs different from the conviction, which is the output of the classification task,The problem that the condition is not satisfied arises.So once(English edition

This problem is solved by applying.

On the contrary, the classification task using the certainty can be diverted to the regression task, and in this case, it is necessary to apply the inverse conversion of the softmax conversion for the same reason as above.

Bias and variance trade-off

In regression, inputxCorresponding toyPredicted value ofIs required to output TheyIt is desirable that it is close to the expected value ofIt is desirable that the variation of is small.However, as shown below, these two requirements are in a trade-off relationship.[14] :

theorem (Bias and variance trade-off) - p(x,y)TheWith the above probability distribution,DTheAs a set of training data selected according to some probability distribution above[Note 3],As a regression algorithmDThe function obtained by training this regression algorithm byAnd the error function is the squared error

Defined byTheDChoose independently of


At this time, the training data set of the prediction errorDExpected value forExpected prediction error[15]



The case of regression was described above, but the same applies to the classification that outputs the certainty.

Bayesian rules

L,p(x,y)Are the loss function and data distribution for supervised learning tasks such as regression and classification, respectively.FForecast lossIt is written as.At this time, the lower limit of the predicted loss

The loss functionLUnderBayesian error(Bayes error) To achieve the lower limitFTheBayesian rules(Bayes rule)[16].. here TheMeasurable functionIn the whole setLower limit.

Bayesian rule is the best predictive function in theory, but in reality it is a probability distributionp(x,y)Is unknownp(x,y)Forecast loss forCannot be calculated and Bayesian rules cannot be obtained.Therefore, known data in supervised learningIt is necessary to search for an algorithm that outputs a value as close to the Bayesian rule as possible.


When the square loss is selected as the loss function, the following theorem holds.[17] :

theorem (Bayes' rule of regression on square loss) - p(x,y)TheWith the above probability distribution,

And.At this time, the generalization errorTo minimizeIt is,

Is. hereE Thep(x,y)Conditional probability distribution determined byRandomly fromyThis is the expected value when you select.

functionTheRegression functionSometimes called[17].


In a classification task (of the type that outputs the class directly rather than the confidence), the Bayesian rule for 0-1 loss is as follows:

Unsupervised learning

Unsupervised learning(unsupervised learning), Unlike supervised learning, the objective variableyIt is not possible to know if there is an equivalent to in the first place.

Unknown probability distribution in unsupervised machine learningVariables that followIs given to the algorithm as training data.The task that the algorithm should solve is the probability distributionAnd somehow learned its important properties,Is to directly estimate the characteristics of[9][18]..A clear "correct answer" unlike supervised learningyThere is no evaluation scale that directly evaluates the validity of the output in unsupervised learning because there is no[18], The judgment of validity becomes subjective[18], Heuristic discussion required[18].

One of the interests of unsupervised learning is the probability density functionEstimate itselfDensity estimationIs the task ofKernel density estimationVarious nonparametric density estimation methods are known in statistics.[18].. HoweverxIf the dimension ofCurse of dimensionalityThis presumption does not work because of[18]Therefore, in many unsupervised learning,With some parametric model ofAttempts to approximate or from training dataAn approach is taken, such as extracting some important property of.

Specific examples are as follows.

Reinforcement learning

Reinforcement learning(Kyogakushu,British: reinforcement learning) Means in an environmentAgentHowever, it is a type of machine learning that deals with the problem of observing the current state and deciding the action to be taken.Agents get rewards from the environment by choosing actions.Reinforcement learning is a measure that gives the most rewards through a series of actions (Policy) To learn.The environment isMarkov decision processIt is formulated as.As a typical methodTD learning,Q learningIt has been known.

  • Reinforcement learning is a method of learning "behavior that maximizes value" through trial and error.
  • Learning is possible even if the correct answer is not known in advance (= teacher data does not exist)
  • There are many application examples in battle games and robots
  • Reinforcement learning using deep learning is called deep reinforcement learning.
  • The name reinforcement learning comes from operant learning, which is a learning mechanism of the brain advocated by Dr. Skinner.
  • Dr. SkinnerSkinner boxThrough a rat experiment called "A reward for a specific movement, that movement is strengthened" was discovered, and this was called operant learning (around 1940).

Other machine learning

For example:

(English edition
It allows you to handle both labeled and unlabeled examples, thereby generating approximate functions or classifiers.
(English edition(Transductive reasoning)
Attempts to predict new output of concrete and fixed (test) cases from the observed concrete (training) cases.
(English edition
Learn about multiple related problems at the same time to improve the prediction accuracy of major problems.

Active learningThe algorithm accesses the desired output (training label) for a limited set of inputs on a budget and optimizes the selection of inputs to obtain the training label. When used interactively, they can be presented to human users for labeling. Reinforcement learning algorithms are fed back in the form of positive or negative reinforcement in a dynamic environment and are used to learn to play games with self-driving cars and human opponents.[19].. Other specialized algorithms in machine learning include computer programsNatural languageThere is topic modeling that gives a set of documents and finds other documents that cover similar topics. Machine learning algorithms are unobservable in density estimation problemsProbability density functionCan be used to determine. Meta-learning algorithms learn their own inductive bias based on past experience. In developmental robotics, robotic learning algorithms generate their own sequences of learning experiences, also known as curriculum, and accumulate new skills through self-guided exploration and social interaction with humans. These robots use guidance mechanisms such as active learning, maturity, motor synergies, and imitation.

Interaction with humans

Some machine learning systems are humanIntuitionIt is trying to eliminate the need for data analysis by humans, but some have incorporated the cooperative interaction between humans and machines. However, the data representation method of the system and the mechanism for exploring the characteristics of the data are designed by human beings, and human intuition cannot be completely excluded.

Relationship with data mining

With machine learningData MiningIs often confused because it has a large intersection and the same technique, but it can be defined as follows.

  • The purpose of machine learning is to make predictions based on "known" features learned from training data.
  • The purpose of data mining is to characterize data that was previously "unknown."発 見It is to be.

The two overlap in many ways. Data mining uses machine learning techniques, but their purpose is often slightly different. Machine learning, on the other hand, also uses data mining techniques as "unsupervised learning" or as a pre-process to improve learner accuracy. The two research areas areECML PKDD With the exception of, basically, academic societies and academic journals are separate. The biggest cause of confusion between them comes from their basic premise. Machine learning evaluates performance based on the ability to regenerate known knowledge, while data mining emphasizes discovering previously "unknown" knowledge. Therefore, "supervised technique" can easily show superior results than "unsupervised technique" when evaluated by known knowledge. However, in typical data mining, training data cannot be prepared, so "supervised technique" cannot be adopted.


An analysis of machine learning algorithms and their performanceTheoretical computer scienceIs a field(English editionIt is called. Learning theories generally cannot guarantee the performance of algorithms because the training examples are finite, while the future is uncertain. Instead, it gives a stochastic range of performance. (English editionby(English editionThere is also an expression called statistical learning theory.[20]

In addition to that, of learningTime complexityI am also studying the feasibility. In computational learning theory,Polynomial timeCalculations ending with are considered feasible.

With machine learningstatisticsAre similar in many respects, but use different terms.

Statistical machine learning

Statistical machine learning is the data of machine learning.Probabilistic generation ruleWhat to learn[21] Refers to.

statistics ThepopulationAnd the specimen, which exists thereProbability distributionIt is a methodology focusing on. In statistical machine learning, we think that data can be obtained stochastically from the population, model the data generation process using a probability distribution, and train the model (or learn the model selection itself) based on the actual data. .. The model of statistical machine learning is also called a generative model / statistical model because it can be interpreted that the data is obtained from the population and the data is generated by sampling from the population.[22].

Specimen-based population (parameter) estimation and selection has long been studied in statistics and there are many theories. Since learning in statistical machine learning is exactly population estimation / selection, the theory of statistics can be applied to machine learning. Various machine learning issues such as learning convergence and generalization performance are being studied using the knowledge system of statistics.

An example of statistical machine learning isneural networkGenerative model in, eg, autoregressive generative net,Variational autoencoder(VAE),Adversarial Generation Network(GAN) and the like. Since data such as images and sounds can be generated by actually sampling from these models (= population), it was studied very well in the latter half of the 2010s, especially in the field of neural networks, and has achieved great results (WaveNet, VQ- VAE-2, BigGAN, etc.).

Mathematical optimization

Many machine learning methods define the error of the model output for the data and update (learn) the parameters so as to minimize the error. The function that calculates the error, that is, the academic system that minimizes the loss function, is used in applied mathematics.Mathematical optimization(The problem to be solved isOptimization problem) Called.

For example,neural networkLet's differentiate the loss functionGradient method(Stochastic gradient descentEtc.), learning is often done. Whether or not the optimization by the gradient method converges to the optimum solution is studied by the theory of mathematical optimization. Also, the constraints imposed on the neural network differ depending on the optimization method used, and all consecutive function applications are differentiable to use the gradient method ()Back propagationIs required (which strongly constrains the sampling of the generative model).


Decision treeLearning
Decision treeThe(English editionIt is a learning used as, and maps observations about an item with conclusions about the target value of that item. As a concrete exampleID3,Random forestThere is.
(English edition
A technique for discovering interesting relationships between variables in large databases.
neural network
Hierarchicalnon-linearA network of transformations.in generalBackpropagation methodLearned at.It has high expressive ability due to non-linearity and is used for various tasks such as classification, regression, and generation.
Genetic programming (GP)
Of living things進化ImitatedEvolutionary algorithmIs a technique based on, performing user-defined tasksProgramTo explore.Genetic algorithmIs an extension and specialization of. By the ability to perform a given taskFitness terrainIt is a machine learning technique that determines and thereby optimizes computer programs.
(English edition (ILP)
Use examples, background knowledge, and hypotheses as uniform expressionsLogic programmingIs a technique for regularizing learning using. Encode a set of known background knowledge and examples into a logical database of facts, with all positive examplesIncluding, Generate a hypothetical logic program that does not contain any negative examples.
Support vector machine (SVM)
Sort,RegressionA series used forSupervised learningIt is a technique. The label of the training example isBinary classification(Classified into two), build a model with a training algorithm and predict which new example will be classified.
Clustering distributes the observed examples to a subset called a cluster, and the distribution is performed according to a pre-instructed standard. The results of clustering differ depending on how a hypothesis (standard) is established for the structure of the data. Hypotheses are defined on a "similarity scale" and are evaluated by "internal compactness" (similarity between members within the same cluster) and distance between different clusters. There are also techniques based on "estimated density" and "graph connectivity". ClusteringUnsupervised learningIt ’s a technique,statisticsOften used in data analysis.
Bayesian network
Random variableFlock and those(English editionTheDirected acyclic graph Expressed in (DAG)Probabilistic graphical modelIs. For example, the relationship between illness and symptoms can be expressed stochastically. If you enter the symptoms in the network, you can output a list of possible diseases with probability. With thisinferenceThere is an efficient algorithm for learning.
(English edition
Unsupervised learningSome of the algorithms try to find a better representation of the input provided during training. As a classic examplePrincipal component analysis,Cluster analysisThere is. There are also algorithms that transform the input into a more convenient representation before classification or prediction, while retaining the information that the input has. At that time, the input can be reconstructed from the unknown probability distribution that the input data follows, but it is not necessary to faithfully reproduce the incredible example in the probability distribution. For example(English editionThe algorithm expresses the input dimension by converting it low under some restrictions.(English editionIn the algorithm, the same expression is converted under the constraint that the input is sparse (many zeros). Of neural networkDeep learningDiscovers multiple levels of representation or hierarchy of features, from low-level extracted features to high-level abstracted features. It is also argued that intelligent machines learn expressions that unravel the potential factors of deviations that explain the observed data.[23].
Extreme learning machine (ELM)
It is a feedforward neural network with one or more hidden layers, and can be applied to classification, regression, and clustering.

Application areas

Machine learning has the following application fields.

2006, online DVD rental companyNetflixOf the companyRecommender systemCompetitions looking for programs that are more than 10% more powerful (more accurately predicting user preferences) Netflix Prize Was held.The competition took several years and the AT & T Labs team called "Pragmatic Chaos."[24] Won the machine learning program in 2009 and won $ 100 million[25].

Actual application

There are:

SortConcrete example
recognition[26]Image recognitionFace recognition[27]
Monitoring work[27]
Inspection / inspection[27]
Organize images[27]
Medical diagnosis[27]
voice recognitionVoice input[28]
Automatic creation of minutes[28]
Call center assistance or alternative[28]
Sentence analysis / sentence recognitionIllegal sentence detection[29]
Understanding needs[29]
Search for similar cases in the past[29]
Anomaly detectionFailure detection[30]
Suspicious behavior detection[30]
analysis[26](Many forecasts[31]Numerical forecastDemand forecast such as sales[32]
Forecast of stock prices and economic indicators[32]
Prediction of time required[32]
Prediction of deterioration[32]
Quality prediction[32]
Prediction of event occurrenceForecast of purchases and cancellations[33]
Failure prediction[33]
Disease prediction[33]
Prediction of compatibility[33]
Coping[26]Behavioral optimizationInventory optimization[34]
Advertising optimization[34]
Campaign optimization[34]
Optimizing store openings[34]
Delivery optimization[34]
Work optimizationself-driving[35]
Robot control[35]
Q & A automation[35]
Expression generation翻 訳[36]
Image generation[36]


Equipped with various machine learning algorithmsSoftware suiteAs,SAS-RapidMiner-LION solver-KNIME-Put-ODM-Shogun toolbox-Orange-Apache mahout-Scikit-learn-mlpy-MCMLL-OpenCV-XGBoost· and so on.

Data Robot[37] There is a method to compare multiple methods by parallel calculation[38].

Academic journals and international conferences


[How to use footnotes]

注 釈

  1. ^ Machine learning and pattern recognition "can be viewed as two facets of the same field."[3]: vii
  2. ^ Because it is often provided by human experts by labeling the training exampleslabelAlso called.
  3. ^ Typicallyp(x,y)Independently according toDSelect each data ofDThe theorem can be proved regardless of the probability distribution selected from


  1. ^ "Machine Learning textbook". www.cs.cmu.edu. 2020/5/28Browse.
  2. ^ (2008) “The Annotation Game: On Turing (1950) on Computing, Machinery, and Intelligence”, in Epstein, Robert; Peters, Grace, The Turing Test Sourcebook: Philosophical and Methodological Issues in the Quest for the Thinking Computer, Kluwer, pp. 23–66, ISBN 9781402067082, http://eprints.ecs.soton.ac.uk/12954/ 
  3. ^ # bishop2006
  4. ^ (1998). “Data Mining and Statistics: What's the connection?”. Computing Science and Statistics 29 (1): 3–9. 
  5. ^ Samuel, Arthur (1959). “Some Studies in Machine Learning Using the Game of Checkers”. IBM Journal of Research and Development 3 (3): 210–229. two:10.1147 / rd.33.0210. 
  6. ^ Mitchell, T. (1997). Machine Learning. McGraw Hill. Pp. 2. ISBN 978-0-07-042807-2 
  7. ^ a b #waterfall p.20.
  8. ^ a b c d e f #ESL p11-12
  9. ^ a b #GBC Verse 5.1.3
  10. ^ #Kanamori p.3.
  11. ^ #waterfall p.8.
  12. ^ a b #waterfall p.36.
  13. ^ #waterfall p.30.
  14. ^ "Lecture 12: Bias-Variance Tradeoff". CS4780 / CS5780: Machine Learning for Intelligent Systems [FALL 2018]. Cornell University. 2020/11/10Browse.
  15. ^ #Kanamori p.13.
  16. ^ #Kanamori p.9.
  17. ^ a b #ESL p22-23
  18. ^ a b c d e f #ESL p559-561
  19. ^ (2006) Pattern Recognition and Machine Learning, Springer, ISBN 978-0-387-31073-2 
  20. ^ Statistical Learning Theory, Takafumi Kanamori, Machine Learning Professional Series, Kodansha, 2015, ISBN 9784061529052
  21. ^ "Statistical Machine Learning Theory and Boltzmann Machine Learning" Muneki Yasuda. Yamagata University
  22. ^ Ueda. "Introduction to Statistical Machine Learning" NII. https://www.youtube.com/watch?v=wqb3k22toFY&t=478
  23. ^ Yoshua Bengio (2009). Learning Deep Architectures for AI. Now Publishers Inc .. p. 1–3. ISBN 978-1-60198-294-0. http://books.google.com/books?id=cq5ewg7FniMC&pg=PA3 
  24. ^ British: Pragmatic Chaos
  25. ^ "BelKor Home Page" research.att.com
  26. ^ a b c # Motohashi 2018 Near the beginning of Chapter 1.3 "Usage of Artificial Intelligence" and "Three Roles of Artificial Intelligence".
  27. ^ a b c d e # Motohashi 2018 Chapter 1.4 "Specific Examples of Recognition" Figure 1-4 "Specific Examples of Image Recognition"
  28. ^ a b c # Motohashi 2018 Chapter 1.4 "Specific Examples of Recognition" Figure 1-5 "Specific Examples of Voice Input"
  29. ^ a b c # Motohashi 2018 Chapter 1.4 "Specific Examples of Recognition" Figure 1-6 "Specific Examples of Sentence Analysis / Sentence Recognition"
  30. ^ a b c # Motohashi 2018 Chapter 1.4 “Specific Examples of Recognition” Figure 1-7 “Specific Examples of Anomaly Detection”
  31. ^ # Motohashi 2018 Chapter 1.5 "What is Analysis?"
  32. ^ a b c d e # Motohashi 2018 Chapter 1.5 "Specific Examples of Analysis" Figure 1-8 "Specific Examples of Numerical Prediction"
  33. ^ a b c d # Motohashi 2018 Chapter 1.5 "Specific Examples of Analysis" Figure 1-9 "Specific Examples of Prediction of Event Occurrence"
  34. ^ a b c d e # Motohashi 2018 Chapter 1.6 “Specific Examples of Coping” Figure 1-10 “Specific Examples of Behavior Optimization”
  35. ^ a b c # Motohashi 2018 Chapter 1.6 "Specific Examples of Countermeasures" Figure 1-12 "Specific Examples of Specific Work"
  36. ^ a b c # Motohashi 2018 Chapter 1.6 “Specific Examples of Countermeasures” Figure 1-13 “Specific Examples of Expression Generation”
  37. ^ British: DataRobot
  38. ^ DataRobot: https://www.datarobot.com


  • Christopher M. Bishop (2006). Pattern Recognition And Machine Learning.Springer-Verlag. ISBN 978-0387310732   (Intermediate and advanced textbooks) →Support page(From here, Chapter 8 "Graphical Models" is available in pdf format)
  • Motohashi, Yosuke (2018/2/15). This book that understands artificial intelligence system projects From planning / development to operation / maintenance (AI & TECHNOLOGY)Shoeisha. ASIN B078JMLVR2. ISBN 978-4798154053  
  • Ian Goodfellow, Yoshua Bengio, Aaron Courville Translation: Hiroo Kurotaki, Shin Kono, Masashi Misono, Jun Hozumi, Naoki Nonaka, Shoji Tomiyama, Takahiro Tsunoda, Supervision: Yusuke Iwasawa, Masahiro Suzuki, Kotaro Nakayama, Yutaka Matsuo / 2018/8). Deep learning (kindle version).Dwango. ASIN B07GQV1X76 
  • Author: Trevor Hastie, Robert Tibshirani, Jerome Friedman, Translation: Masaru Sugiyama, Tsuyoshi Ide, Toshihiro Kamishima, Takio Kurita, Eisaku Maeda, Yoshihisa Ijiri, Toshiharu Iwata, Takafumi Kanamori, Atsushi Kanemura, Masayuki Karasuyama, Yoshinobu Kawahara, Shogo Kimura, Yoshinori Konishi, Tomoya Sakai, Daiji Suzuki, Ichiro Takeuchi, Toru Tamaki, Daisuke Deguchi, Ryota Tomioka, Hitoshi Habe, Shinichi Maeda, Daichi Mochihashi, Makoto Yamada (2014/6/25). Basics of Statistical Learning-Data Mining, Inference, PredictionKyoritsu Shuppan. ISBN 978-4320123625  
  • Masato Taki (2017/10/21). This is an introduction to deep learning.KS Information Science Specialized Book Machine Learning Startup Series. Kodansha. ISBN 978-4061538283  
  • Takafumi Kanamori (2015/8/8). Statistical learning theory.KS Information Science Specialized Book Machine Learning Startup Series. Kodansha. ISBN 978-4061529052  
  • Yasuaki Ariga, Shinta Nakayama, Takashi Nishibayashi, "Machine Learning Beginning at Work", January 2018, 1.ISBN 978-4-87311-825-3.

Further reading

Related item

外部 リンク


Back to Top