Today, AI has become a very familiar entity, but there are still many people who say that they “somehow understand” without knowing its essential meaning and mechanism.
What is AI (artificial intelligence)?
First, what is “AI”? Let’s review the meaning and definition of the word.
The Birth of the Word AI (Artificial Intelligence)
The term AI was first used in 1956. It was proposed by Computer Scientist and Cognitive Scientist Professor John McCarthy at the Dartmouth Conference held at Dartmouth College in the United States.
What does “AI” stand for?
AI is an abbreviation for Artificial Intelligence. Artificial means “artificial” and Intelligence means “intelligence”.
Antonym of AI (artificial intelligence)
The antonym of AI is Natural Intelligence. The abbreviation is NI. The Japanese translation of “Natural Intelligence” is the word “Natural Intelligence”, which refers to the intelligence created by nature such as humans and animals.
Definition of AI (artificial intelligence)
The Japanese Society for Artificial Intelligence translates the words of Professor John McCarthy, the creator of the word AI, as “the science and technology of creating intelligent machines, especially intelligent computer programs . ”
However, as research on AI progresses, the current situation is that different researchers use different terms to define it. The definition of AI by major researchers in Japan is as follows.
▪️Definition of artificial intelligence (AI) by leading researchers in Japan
| researcher | Belongs | definition |
|---|---|---|
| Hideyuki Nakashima | President , Future University Hakodate |
An artificially created entity with intelligence. Or it is a field that studies intelligence itself by trying to create it. |
| Toyoaki Nishida | Professor, Graduate School of Informatics, Kyoto University | “Mecha with intelligence” or “Mecha with heart” |
| Riichiro Mizoguchi | Hokuriku Advanced Science and Technology Graduate University Professor |
It is an artificially created system that behaves intelligently. |
| Makoto Nagao | Professor Emeritus, Kyoto University Former Director of the National Diet Library |
It is a system that simulates human brain activity to the limit. |
| Koichi Hori | Professor, Graduate School of Engineering, The University of Tokyo |
It is a new world of artificially created intelligence. |
| Minoru Asada | Professor, Graduate School of Engineering, Osaka University | Artificial intelligence cannot be defined clearly because the definition of intelligence is not clear |
| Hitoshi Matsubara | Future University Hakodate Professor |
Artificial intelligence that is ultimately indistinguishable from humans |
| Hideaki Takeda | National Institute of Informatics Professor |
An artificially created entity with intelligence. Alternatively, it is a field that researches intelligence itself by trying to create it (same as Mr. Nakajima) |
| Takashi Ikegami | Professor, Graduate School of Arts and Sciences, The University of Tokyo |
Artificial intelligence is defined as a system that can artificially create emotional and humorous interactions, such as we naturally come into contact with pets and people, regardless of or against the laws of physics. A system that makes you want to understand discourse through conversation and socializing, rather than wanting to understand analytically. that’s artificial intelligence |
| Takahira Yamaguchi | Professor , Faculty of Science and Technology, Keio University | Constructive system to imitate, support, and transcend human intellectual behavior |
| Satoshi Kurihara | Professor, Graduate School of Information Systems, The University of Electro-Communications | It is an intelligence created by engineering, but the level of that intelligence is imagining something that exceeds that of humans. |
| Hiroshi Yamakawa | Director of Dwango Artificial Intelligence Laboratory | Among computer intelligences, I think that the case where humans design directly or indirectly can be called artificial intelligence. |
| Yutaka Matsuo | Associate Professor, Graduate School of Engineering, The University of Tokyo |
Human-like intelligence created artificially, or the technology to create it |
It is not generally wrong to think of ” What is AI?” The current situation is that it is not defined .
◎What is AI?
・Proposed by computer scientist Professor John McCarthy at the Dartmouth Conference in 1956
・”Science and Technology for Creating Intelligent Machines, Especially Intelligent Computer Programs” / Japanese Society for Artificial Intelligence
・It’s like “a computer with human-like intelligence”, but there’s no clear definition.
History of AI (artificial intelligence) research
Currently, AI is being accepted by human society as a familiar existence, such as voice assistants such as Apple’s “Siri”, iRobot’s cleaning robot “Rumba”, and Softbank’s emotion recognition humanoid robot “Pepper”.
The progress of AI technology in recent years has been remarkable, but AI research began to flourish around the late 1950s. AI has gradually evolved through booms and winter periods (stagnation). Here, we will briefly introduce the history of such AI research.
First AI boom (late 1950s to 1960s)
The first AI boom took place in the late 1950s and 1960s, when the Dartmouth Conference that coined the word “AI” was held.
Behind the boom was the ability to “infer” and “search” by computer, and it became possible to derive answers to specific problems . However, AI at that time was limited to problems with clear rules and definitions.
The AI boom gradually died down as it became clear that it was not possible to solve problems in which various factors that occur in the real world are intricately intertwined. In the 1970s, we entered a period of winter (stagnation).
Second AI Boom (1980s-1990s)
The second AI boom arrived in the 1980s and 1990s. The birth of the ” expert system ” made “knowledge expression” possible, which is a major factor.
An expert system is an artificial intelligence whose knowledge is composed of a set of rules such as “If yes, do x. Otherwise, do △.” There is no mechanism for self-learning, but it works by having experts anticipate every conceivable situation and prepare countermeasures and judgments in advance.
The more rules there are, the higher the accuracy will be, but all the necessary information must be understood by the computer by human hands, and what can actually be used is limited to information in specific areas. Only. Since the amount of knowledge that could be used in this way seemed to be limited, from around 1995, AI entered the winter era again.
Third AI Boom (2000s to present)
The third AI boom began in the 2000s, and it is still in the midst of it as of 2019 when this article was written.
The reason for the boom was the progress in the practical application of ” machine learning ,” in which AI itself acquires knowledge from large amounts of data (big data) . Furthermore, in 2006, deep learning , in which AI learns the elements (features) that define knowledge by itself , was proposed, spurring the boom.
Representative Algorithms of AI (Artificial Intelligence)
We will explain representative algorithms such as “neural networks”, “genetic algorithms”, and “expert systems”, which are also called AI three families and play an important role in AI research.
Neural network
Neural networks are AI modeled on neurons (nerve cells in the brain)
A neural network is an AI modeled on the structure and function of neurons (the nerve cells that make up the brain of living organisms) . When a neuron receives an electrical signal above a certain value from another neuron, it gets excited and sends an electrical signal to the connected neuron. This is a numerical model of the mechanism of such cooperative behavior between neurons.
A neural network consists of an input layer that receives data, an intermediate layer (hidden layer) that processes the weights flowing from the input layer, and an output layer that outputs results. When a human becomes a teacher and teaches a set of example questions and model answers (teacher signal) to the neural network, the neural network itself will make judgments and inferences even in areas that have not been taught .
genetic algorithm
Genetic algorithm is AI (artificial intelligence) with Darwin’s theory of evolution as a motif
A genetic algorithm is an AI based on Darwin’s theory of evolution.
Darwin’s theory of evolution can be summarized as follows.
In living things, only superior individuals can leave offspring according to the environment, and inferior individuals are eliminated. In addition, individuals may undergo mutations, and rarely become excellent individuals . It has evolved over and over again.
(Source) Yoichiro Miyake and Yukihito Morikawa “Artificial Intelligence through Pictures” (SB Creative) p.62
Genetic algorithms try to derive the optimal solution using evolutionary methods as this “excellent individual” = “good answer” .
Genetic algorithms are best at finding the best answer among a huge number of combinations. You can quickly find the optimal solution for problems that cause combinatorial explosions that are difficult to calculate manually.
Expert system
Expert systems are AI (artificial intelligence) modeled on human thinking
An expert system is an AI that is modeled after the human way of thinking. It differs from other AI models in that it has no self-learning mechanism .
First, we hear from specific experts (experts) about conceivable situations and how to deal with them, judgments, and predictions, and then define rules based on that. Based on the rules defined there, it determines which situation the user’s inquiry applies to, and makes the defined judgments and predictions.
They are especially active in diagnosing diseases in the medical field, for example:
★Example: Diagnosis by expert system
Answer: ①I have a fever ②I have a runny nose ③I have a cough
Rule 1: If you have a fever, judge it as food poisoning
Rule 2: If you have a runny nose, you have a cold
Rule 3: If you cough, you have tuberculosis
(Source) Yoichiro Miyake and Yukihito Morikawa “Artificial Intelligence through Pictures” (SB Creative) p.73
As described above, a diagnosis prepared in advance is made according to the answer from the user.
The more rules there are, the more accurate it will be, but if there are too many rules, it may be difficult to maintain the consistency of each rule. Furthermore, if there are omissions or omissions in important rules, it will be impossible to make correct decisions.
In addition, it is a concern of the expert system that setting the rules requires the help of experts and even if the rules are set correctly, it is not possible to derive answers that are better than experts.
AI (artificial intelligence) learning method ① machine learning
Machine learning is “learning” in AI . It has the meaning of “the machine itself learns” like a human learns.
In other words, the standard is to be able to do more than what was programmed by the programmer .
Machine learning can be broadly classified into three categories: supervised learning, unsupervised learning, and reinforcement learning .
AI (artificial intelligence) learning method ② Deep learning
Deep learning is a machine learning method that uses multilayered neural networks . As long as there is enough training data, the neural network itself can automatically extract the features of the data group.
Singularity (technical singularity) is a term that has been used among AI researchers since the 1980s, and refers to the critical point between humans and artificial intelligence . In other words, it represents the time when AI at the same level as the human brain will be born.
Please refer to this article for details
In recent years, we have welcomed the fact that AI -based technology has become more familiar and the world has become more convenient. have more opportunities to
However, it is also true that there is still an unfillable gap between AI (Natural Intelligence) that humans have and AI that is expressed on computers.
In fact, most of the AI currently being developed is problem-specific, and the current situation is that it functions only to solve a single modeled/mathematicalized problem .
AI-powered functions
Using the learning methods introduced so far, AI is currently active in various fields.
We have summarized the functions that utilize AI by category, so let’s check what they are used for.
▪️Examples of functions utilizing AI by category
| Category | Examples of application areas |
|---|---|
| computer vision | Image classification/image generation/object detection |
| natural language processing | Machine translation/Language modeling/Answering questions |
| medical care | medical image segmentation |
| methodology | Distributed representation (word embedding) / representation learning |
| game | video games/board games |
| Graph | Link prediction/node classification |
| speech | Speech recognition/speech synthesis |
| Time series | Time series classification/imputation |
| audio | Music generation/audio classification |
| robot | Calibration/self-localization |
| music | Music information retrieval/music modeling |
| computer code | Dimensionality Reduction/Program Synthesis |
| inference | Decision Making/Common Reasoning |
| knowledge base | Knowledge Graph/Causal Discovery |
| hostility | Attack/Defense/Hostile Text |
| others | Recommendation/Topic model |
The above is just one example. As the practical application of AI technology is progressing day by day, it is expected to be used in a wider range of applications in the future.
Monster Lab AI Development Case
At Monster Lab, we support digital products that utilize AI from each stage of planning, design, development, and operation.
Here, we will introduce the AI development case of Monster Lab.
Automatic measurement app that utilizes AI image recognition (Unimate Co., Ltd.)
In the conventional rental uniform business, size discrepancies due to human error frequently occur. It was a big problem that a large amount of cost was generated for returns and exchanges, including labor.
To solve the problem, the company developed an automatic measurement PWA “AI x R Tailor” that utilizes AI image recognition. Monster Lab participated from the planning stage and was in charge of all product development processes.
We succeeded in reducing not only the client’s work cost for measurements, but also our own cost for returns.
Mainichi Broadcasting System (MBS) | Efficient video editing work through analysis and adjustment using AI
Analysis and adjustments using image recognition AI and audio signal processing contribute to a 40% reduction in editing work time
Mainichi Broadcasting System (MBS) is a television broadcasting station covering the Kinki wide area.
At the “39th Suntory 10,000 People 9th” held by the company, Monster Lab analyzed and adjusted 14,215 general posted videos using image recognition AI and audio signal processing. By streamlining the initial work of video editing, MBS editing work time was reduced by 40% compared to the previous year.
Summary: AI (artificial intelligence) holds the key to DX
AI (artificial intelligence) holds the key to DX
How was the basic knowledge about AI, such as the meaning and definition of words, the history of development, and learning methods?
Regarding the evolution of AI, some people may feel uneasy and fearful about the part that “it may exceed human ability”, but now that technology continues to advance rapidly, it is no longer a road to avoid. It would not be an exaggeration to say that they are gone.
Human society is constantly changing with the times. There is no doubt that the wave of DX will continue to surge in various industries and fields due to the functions that AI will bring.
The important thing is to think positively about how AI will enrich people’s lives. It is more constructive to have a correct understanding of AI and think about how a new society should be.
For companies, the key to medium- to long-term growth strategies will be to quickly incorporate AI functions from the perspective of transforming existing businesses and creating new businesses and promoting DX.
