In this article, we will introduce AI levels. There are various types of AI around us, from simple to complex, and they are categorized as “levels.”
It deals with interesting topics, such as what level the AI products around us belong to, or whether it is possible for higher-level AI to be born.
Table of Contents
- What is AI
- AI has levels
- Explanation of AI levels 1 to 4
- Level 1: Simple control algorithm
- Level 2: Rule-based reasoning program
- Level 3: Machine Learning
- Level 4: Deep learning (deep learning)
- History of AI levels
- 1st AI boom
- Second AI boom
- Third AI boom
- in conclusion
What is AI
The term AI ( artificial intelligence ) has many meanings and definitions. Simply put, AI is a technology that “realizes intelligence using the method of calculation.”
When describing AI, it is sometimes more intuitively referred to as an “intelligent computer.” The question I want to ask in this article is what is that “intelligence”?
AI has levels
AI has levels according to its intelligence ability and what it can do. There are categories from level 1 to 4, and the higher the number, the more complicated the mechanism.
In the next chapter, we will introduce each level of AI in detail.
Explanation of AI levels 1 to 4
AI can be divided into four levels according to the processing mechanism and the algorithms it employs. The levels are as follows.
- Level 1｜Simple control algorithm
- Level 2｜Rule-based reasoning program
- Level 3｜Machine Learning
- Level 4｜Deep Learning
I will explain each.
Level 1: Simple control algorithm
First, I will introduce the simplest AI, “Level 1”. Level 1 AI is a control program so simple that it can no longer be called AI from the current AI technology.
A control program consists of conditional branches and usually produces simple outputs depending on the input. From this, it can be said that AI research started from a field of control engineering research.
Practical example of Level 1
Level 1 AI is all around us. Have you ever heard the term AI-equipped home appliances? When we talk about “AI-equipped” in consumer electronics, it usually refers to this Level 1 AI.
For example, AI-equipped refrigerators automatically control cooling according to the type of food stored and the lifestyle patterns of the user. In addition, the AI-equipped air conditioner controls the air volume and set temperature according to the user’s position and the outside weather.
Level 2: Rule-based reasoning program
Level 2 AI goes one step further than Level 1 AI and refers to AI that selects the next behavior from among multiple behavioral patterns according to input.
AI that behaves in this way is called rule-based AI. Pre-designed behavioral patterns are accumulated as ‘knowledge’, and level 2 AI makes inferences and searches according to input.
In this sense, Level 2 AI appears to think by itself, so it is close to the general image of AI ( artificial intelligence ).
Practical example of Level 2
A typical example of Level 2 AI is a question and answer system. A question and answer system is a system that allows questions and answers to be exchanged in natural languages such as Japanese and English.
The question-and-answer system became a hot topic when “Watson,” developed by IBM, won the American quiz show Jeopardy!.
Level 3: Machine Learning
Level 3 AI is AI that learns through a mechanism called machine learning.
Level 1 and level 2 AI have rules set by humans to output a fixed output according to the input, but level 3 AI finds the pattern of data by itself and adjusts the output according to the input. .
Level 3 AI requires humans to set the feature values of the data that the AI refers to during learning. Therefore, it does not learn completely autonomously, but it can produce much more diverse output than Level 2 AI.
Practical example of Level 3
A prime example of Level 3 AI is a search engine. When we enter a keyword in the search window and perform a search, the articles of the search results are displayed collectively.
The AI installed in the search engine determines which articles should be displayed at the top and which articles should not be displayed.
Search engine AI adjusts and displays search results that meet user needs while referring to a huge data group called big data. This technology uses machine learning.
Level 4: Deep learning (deep learning)
Level 4 AI is an AI that learns through a mechanism called deep learning. The major difference from Level 3 AI is that it can carry out the learning process, including adjustment of feature values, by itself.
Deep learning can process much more data than machine learning. As a result, many AI applications that were technically impossible in the past have been successfully realized.
Practical example of Level 4
A typical example of Level 4 AI is self-driving technology. Autonomous driving is a technology in which a machine controls a vehicle by itself without human intervention. It is hoped that this will be realized in the near future.
AI used in autonomous driving technology predicts the movements of cars and people in front of you, and also instantly determines the type of obstacles in front of you as you continue driving.
For that purpose, it is necessary to calculate a large amount of data in an instant, which has been considered difficult with conventional machine learning, but with the development of deep learning technology, it is partially practical.
History of AI levels
Next, I will introduce the history of AI levels. In the history of AI, there have been booms with every innovation.
The following three AI booms have occurred so far.
- 1st AI boom
- Second AI boom
- Third AI boom
In this chapter, we will explain the relationship between the above boom and AI level.
1st AI boom
The simplest level 1 AI was conceived in the early days of AI. It was during this first AI boom that the name AI was coined.
During the first AI boom of the 1950s and 1960s, research into simple inference and search programs was extensive. However, we soon hit the limits of technology.
It was difficult to apply AI, which can only solve so-called “toy problems,” to society.
Second AI boom
The invention of Level 2 AI led to the second AI boom. In the 1980s, AI called ” expert system ” was actively developed.
An expert system is a system that behaves like an expert by comprehensively inputting specialized knowledge such as medicine and law.
However, expert systems have the drawback of not being able to deal with unknown data. Level 2 AI finds it difficult to do things other than what it has been taught in advance, and cannot deal with the ambiguity that humans have.
Thus, the second AI boom came to an end. However, this does not mean that research into AI has ceased to be active. We made steady progress in the development of technologies such as Level 3 AI that allow AI to learn autonomously and deal with unknown data.
Third AI boom
And the breakthrough of deep learning used for level 4 AI sparked the 3rd AI boom.
Generally, this breakthrough is said to have started when a research team using a deep learning model won the 2012 world competition for image recognition with overwhelming results.
The third AI boom has greatly expanded the range of technologies that use AI, such as image recognition, voice recognition, and natural language processing.
In addition, shocking technologies that have never been thought of before have been developed, such as AI for shogi and Go, which surpasses human ability, and automatic driving technology.
in conclusion
In this article, we introduced the levels of AI and their history. You can see that the AI around us ranges from very simple mechanisms to complex ones such as deep learning.
Also, if you look at how level 1-4 AI contributes to the AI boom, you’ll find a lot of interesting things.
