In recent years, AI has been used in various services and machines! I’m sure you’ve seen or heard about AI at least once.
However, few people know how it works.
Let’s take a look at what AI is and how it works! It also mentions how AI is created, so if you are interested in AI creators or AI engineers, please read it!
Contents [ Hide ]
- What is AI?
- How AI works
- What can AI do? Introducing 5 cases together!
- How to make AI 4 steps
- 3 ways to learn AI programming
What is AI?
AI is an abbreviation for Artificial Intelligence, and is expressed as artificial intelligence in Japanese.
In fact, the definition of AI is extremely vague. It is defined in different terms by different researchers.
It is also discussed in various fields such as computer science, cognitive science, medicine, and psychology, and there is currently no unique definition.
Let’s take a look at the definition of AI by major researchers in Japan.
|Hideyuki Nakashima(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(Kyoto University University)||“Mecha with intelligence” or “Mecha with heart”|
|Riichiro Mizoguchi(Hokuriku Advanced Institute of Science and Technology)||It is an artificially created system that behaves intelligently .|
|Makoto Nagao(Kyoto University)||It is a system that simulates human brain activity to the limit.|
|Koichi Hori(University of Tokyo)||It is a new world of artificially created intelligence.|
|Minoru Asada(Osaka University)||Artificial intelligence cannot be defined clearly because the definition of intelligence is not clear|
|Hitoshi Matsubara(Future University Hakodate)||Artificial intelligence that is ultimately indistinguishable from humans|
|Hideaki Takeda(National Institute of Informatics)||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(University of Tokyo)||A system that can artificially create interactions full of emotion and jokes that we naturally come into contact with pets and people, regardless of or against physical laws.|
|Takahira Yamaguchi(Keio University)||Constructive system to imitate, support, and transcend human intellectual behavior|
|Satoshi Kurihara(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(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(University of Tokyo)||Human-like intelligence created artificially, or the technology to create it|
▼ Click here for more information about AI
AI specialized news media AINOW
What is artificial intelligence (AI)? Thorough explanation of definitions and history! The latest AI is too dangerous.https://ainow.ai/artificial-intelligence-3/The topic of AI (artificial intelligence) has been appearing in the news almost every day. According to the “2020 Artificial Intelligence Business Survey” by Fuji Chimera Research Institute, the domestic AI business market size in 2020 is expected to expand by 15.4% from the previous year to 1,108.4 billion yen. Also, after 2021, DX (de…
general-purpose and specialized
There are two categories of AI: “specialized type” and “general-purpose type”.
Refers to artificial intelligence that automatically learns and processes tasks in a specific domain. Artificial intelligence with technologies such as image recognition, voice recognition, and natural language processing.
All AI currently in use falls under this category. In recent years, with the development of a machine learning method called deep learning, AI programs with extremely high accuracy have been created, and the scope of their activities is expanding.
It refers to artificial intelligence that can handle various tasks in the same way as humans, rather than dealing only with specific tasks.
General artificial intelligence is expected to become an AI that can increase its knowledge and work on problem solving by giving it the autonomy to learn on its own in any field.
However, it is still in the research stage and has not yet been realized.
Weak AI and Strong AI
There are other ways to classify AI as “ weak AI ” and “ strong AI ”.
A weak AI is one that replaces only a portion of human intelligence and only handles specific tasks.
On the other hand, since it does not have all the cognitive abilities of humans, it cannot respond to unexpected situations. Any AI that is currently in the practical stage falls under the category of weak AI.
It refers to AI that has all cognitive abilities like humans. A strong AI can think and act like a human.
It is also expected to be able to “learn and process unexpected situations based on past experiences.”
Summary of categories of AI
From the above, specialized type ≒ weak AI , general purpose type ≒ strong AI. Both of these two classifications are based on comparisons with humans.
“Specialized/general-purpose” is a classification that focuses on the breadth of problem-solving targets, and ” weak AI / strong AI ” is a classification that focuses on thinking ability.
If you want to know more about the types of AI, click here>>
How AI works
While learning about AI, you will often hear the terms machine learning and deep learning.
By arranging their relationships,
AI⊃Machine Learning⊃Deep Learning _
An inclusion relationship is established. It will look like the following figure.
How machine learning works
It is an algorithm that automatically improves the accuracy of predictions for unknown data while learning input data, and is one of the AI technologies.
In other words, it is a method of reading data into a computer and analyzing it based on an algorithm.
By repeatedly learning case data, it discovers hidden features and patterns. Then, by applying the found features to new data, new data analysis and predictions are made.
Machine learning analyzes mainly based on the following three types of algorithms.
▼ Click here to learn more about machine learning
AI specialized news media AINOW
What is machine learning? Easy-to-understand explanations from definitions to learning methods, five algorithms, and application exampleshttps://ainow.ai/2019/11/26/180809/Hello, this is Satoshi, an AINOW intern. In this article, we will explain machine learning, which is often confused with AI and deep learning, in an easy-to-understand manner so that you can understand their relationships and differences. Also, essential terms for understanding machine learning (supervised learning, unsupervised learning, each al…
(1) Supervised learning
Supervised learning is a method of learning with the correct answers given to the training data.
Supervised learning uses teacher data as known information for learning, and constructs regression models and classification models that can deal with unknown information.
A computer is trained on a large number of photos pre-labeled as “dog” or “cat” (teaching data) to build a model.
Given an unlabeled photo, the model will output the label attached to the trained image with the closest features to that image. In other words, the model will be able to detect that a given image is a “dog” or a “cat”.
(2) Unsupervised learning
Unsupervised learning is a method of learning without giving correct answers to the training data.
It is mainly used for grouping data.
Even if there are no labels in the training data, training a computer with a large number of images enables grouping and summarization of information based on image features (eg size, color, shape).
③ Reinforcement learning
Reinforcement learning does not learn from given data, but rather takes action on its own and, through trial and error, learns actions and choices that provide rewards (evaluations).
An example of reinforcement learning is dog training. If the dog touches you, give him a treat. If the dog repeats the “Hand” trial, the dog will learn that “handing ⇒ getting food”.
Taking the walking of a robot as an example, if the “distance walked” is the reward, “extending the walking distance by trial and error in how to move the hands and feet” corresponds to reinforcement learning.
Click here to learn more about Reinforcement Learning>>
How deep learning works
The basis of deep learning is a structure called a neural network.
A neural network is a model that mimics the structure of neural circuits in the brain. A neural network consists of three layers: an input layer, an intermediate layer (hidden layer), and an output layer.
In the middle layer (hidden layer), the data received from the previous layer is “weighted” and “transformed” and passed to the next layer.
The output of the neural network is compared with teacher data, etc., and the weighting method is adjusted so that the degree of matching is higher.
Learning using a neural network with two or more intermediate layers (hidden layers) is called deep learning.
Multiple hidden layers can produce a more complex output that matches better than a single hidden layer.
The image below is an image diagram.
Deep learning has made it possible for computers to detect features by themselves.
In other words, by giving a large amount of training data, the computer can learn by itself.
What can AI do? Introducing 5 cases together!
In what specific fields is AI active? Here are 5 examples!
- Disease diagnosis support (image recognition)
- Smart speaker (voice recognition, natural language processing)
- Games (exploration and reasoning)
- Automation of warehouse operations (AI x robot)
- Stock price forecast (data analysis/forecast)
(1) Disease diagnosis support (image recognition)
Image recognition technology is expected in the medical field. Conventionally, the role of doctors is to detect diseases from images such as X-rays and CT scans.
However, human error can occur, such as inconsistencies in judgments due to human visual observation alone. AI is starting to be used to solve these problems.
For example, deep learning can be used to detect cancer cells from images with high accuracy. It is possible to identify detailed features that were difficult to distinguish with conventional human eyes.
If this is put into practical use, it will help doctors make diagnoses and improve the quality of medical care.
(2) Smart speaker (voice recognition, natural language processing)
Do you know “Alexa”?
Alexa is a smart speaker developed by Amazon. Utilizing voice recognition of artificial intelligence, etc., all operations are automated with voice instructions.
It is already used by many people and is helping to realize a comfortable and stress-free life.
③ Game ( exploration and reasoning )
By using AI technology, it is possible to instantly classify patterns by using search and inference to find items that meet conditions from a large amount of complex data collection.
Using these technologies, it is now possible to play against AI in games such as shogi and chess. Nowadays, it is impossible for a professional shogi player to beat AI. It has evolved so much.
④Automation of warehouse operations (AI x robot)
In the logistics industry, AI has significantly improved operational efficiency, and each process such as warehousing and shipping, order processing, inspection work, and parcel sorting has begun to be automated.
For example, in 2018 Fast Retailing, which owns UNIQLO, collaborated with Daifuku to introduce and automate warehouse operations using AI.
In this case, we succeeded in reducing the number of employees from about 100 to 10.
(5) Stock price prediction (data analysis/prediction)
Monex, Inc. offers “AI Stock Navi” for stock price prediction. By learning the feature values of past stock price movements, AI predicts stock prices one month ahead for each stock.
AI has the potential to discover features that humans cannot, so we can expect highly accurate analysis.
How to make AI 4 steps
Now, how is AI actually created? Here, the explanation is divided into 4 steps.
- Preprocessing of collected data
- Repeat ① to ③
(1) Preprocessing of collected data
The first step to create artificial intelligence is to collect and preprocess data for machine learning.
Data preprocessing is the process of shaping data into a usable form. This is an important step as the quantity and quality of this data has a large impact on the learning outcome.
There are various types of data used in AI learning, such as text data, image data, and voice data.
Those data often contain challenges.
For example, the data may have outliers or missing values, the data format may be inconsistent, or it may not be suitable for AI learning.
For these, outliers and missing values are deleted, data cleansing is performed, and data conversion and normalization are performed to process the data into data that is easy for AI to learn. Labeling is also a preprocessing step for this data.
Next is the design of a “model” that determines how to interpret and analyze the input data and perform learning.
Since the model suitable for each characteristic of AI to be developed is different, we will design an appropriate model. Then, machine learning is actually performed based on the input data.
Learning is not the end of the story. Evaluate whether learning has achieved the desired level of analytical competence. In other words, we need to check the accuracy of AI.
First, let AI analyze the data prepared for testing. Then check to see if the results are what you expected or not.
If the target results are not obtained in (3), the training data will be reexamined or added.
Repeat this until you get the accuracy. And if you get the expected result, you’re do
3 ways to learn AI
So far, I have explained how to make AI in 3 steps. Next, I will introduce the study method of AI programming for those who actually want to study AI programming and create AI.
- learn at programming school
- learn online
- learn from books
① Learn at a programming school
The first is to learn at a programming school. Self-study is recommended for those who have high hurdles and want to create a learning pace.
The programming school provides generous support, and there are many ingenuity to learn until the end.
There are also schools that provide support for finding employment and changing jobs, so it is recommended for those who want to study with a view to finding a job or changing jobs.
② Learn in online courses
The second is an online course. It is recommended for those who find it difficult to learn from books alone and want to use their time efficiently.
There are also online courses where you can take lectures by famous instructors and ask questions.
Also, since the financial burden is less than attending a programming school, it is recommended for those who are concerned about the cost.
③ Learn from books
The third is books. To be honest, it’s more difficult than the above two, but if you’re good at self-study, you can learn AI programming from books.
In studying programming, you cannot avoid fighting with repeated compilation errors. This is a point where many people stumble when learning programming on their own.
In such a case, a site that provides a question service that specializes in the worries of IT engineers is useful. A particularly recommended site is teratail .
We recommend that you study a combination of such sites and books.
|cost||-¥ 10,000||Tens of thousands to hundreds of thousands of yen||¥100,000-¥1,000,000|
|period||Three months-||1 month – 6 months||1-3 months|
|Degree of recommendation||☆☆||☆☆☆☆||☆☆☆|
This time, I explained the basic information of AI and the mechanism of AI.
AI has several classifications and various methods! I hope that by grasping the whole picture of AI, it will serve as a reference when utilizing AI in the future!