On the Classification and Application of AI Technology
AI is already being utilized in every aspect of our lives, making it difficult to define exactly what AI is. Generally, AI refers to technologies and academic fields that enable intelligent processing and problem-solving similar to that of humans, aiming to automate various tasks and functions, and to mimic or even surpass human intelligence.
Some scholars and companies providing AI-specialized services sometimes argue that 'it is impossible for AI to surpass humans' or 'AI must never surpass humans.' However, such discussions often become incoherent if it is not clear which AI is being referred to.
AI's ability to control everything from home appliances to the autopilot systems of racing cars and airplanes, and even fighter jets, missiles, and precision machinery, already far exceeds human capability. Furthermore, even automated translation software far surpasses the English proficiency of the average Japanese person.
Under limited conditions, such as whether a scholar majoring in English literature or a translation AI can translate Shakespeare's works into Japanese more accurately, there are aspects where one could say human translation is still more accurate. However, if you ask whether an English literature scholar who has spent their entire life studying only Shakespeare can translate medical terminology, nanotechnology terminology, or informatics terminology more accurately than a translation AI, it is not necessarily the case that an English literature scholar without this specialized knowledge can produce a more accurate translation than the AI. As such, the reality is that depending on the field of application, AI's capabilities already far exceed those of humans.
The debate over whether AI will surpass human capabilities is particularly related to #Artificial General Intelligence (#AGI: Artificial General Intelligence). AGI refers to AI that can cover a broad range of human intelligence and adaptively respond to any task. Therefore, whether or not AGI can transcend human capabilities is a central focus of current technological development and research. Many people do not understand the difference between AGI and #Artificial Super Intelligence (#ASI: Artificial Super Intelligence), but the definition of ASI is one focus of the debate. From a philosophical perspective, from the era of #Ancient Greek Philosophy to the present day, the #definition of intelligence itself has been a subject of debate. Therefore, clarifying the definitions of AGI and ASI is an extremely difficult problem.
Therefore, in this chapter, to make it easier to understand which AI is being discussed, I have classified the various properties of AI from the following perspectives.
(1) Classification by Application Domain
Classified by the domain or industry where AI technology is applied. Examples include medical AI, autonomous driving AI, and financial AI.
(2) Classification of Natural Language Processing AI
Classified by AI that is specifically specialized in natural language processing among AI technologies. Examples include machine translation, text generation, and sentiment analysis.
(3) Classification by AI Application Area
Classified by the applications or services where AI technology is applied. Examples include voice assistants, chatbots, and image recognition software.
(4) Classification by AI Task
Classified by the type of task the AI solves. Examples include classification, prediction, optimization, and generation.
(5) Classification by AI Technology
Classified by the algorithms or methods that form the basis of AI technology. Examples include machine learning, deep learning, and reinforcement learning.
(6) Classification by Approach
Classified by the approach or philosophy in the development and research of AI technology. Examples include symbolic AI, connectionism, and evolutionary algorithms.
Through these classifications, we can understand the various aspects and properties of AI and provide a basic framework for advancing discussions. Discussions regarding the application fields and performance of each AI should be clarified based on such classifications.