AI and Machine Learning Basics | Key IT Passport Exam Terminology

The Relationship Between AI, Machine Learning, and Deep Learning

AI (Artificial Intelligence) is a broad concept encompassing all technologies that mimic human intelligence. Machine learning is a method for automatically learning patterns from data, and deep learning is a specialized technique within machine learning that uses multi-layer neural networks. These have an inclusive relationship, with the hierarchy narrowing in the order AI ⊃ Machine Learning ⊃ Deep Learning.

The Three Categories of Machine Learning

Supervised Learning

Supervised learning is a method that learns using pairs of input data and correct answer labels. Typical examples include image classification, spam detection, and price prediction. Main techniques include regression for numerical prediction and classification for determining categories.

Unsupervised Learning

Unsupervised learning discovers the structure of data without using correct answer labels. Clustering and dimensionality reduction are typical examples, used for customer segmentation, anomaly detection, and more.

Reinforcement Learning

Reinforcement learning is a method that learns actions to maximize rewards through repeated trial and error. Typical examples include the Go AI AlphaGo, robot control, and game AI.

Generative AI and LLMs

Generative AI is a general term for AI that newly generates text, images, audio, and more. LLMs (Large Language Models) are language models trained on vast amounts of text, with ChatGPT, Claude, and Gemini being well-known examples. On the exam, terms like prompt, hallucination (false information generation), and fine-tuning are frequently tested.

Key Points on the IT Passport Exam

In the past exam questions for fiscal year 2025 (Reiwa 7), AI-related problems accounted for 7 questions, making it a frequently tested area. The exam tests identification of terms for the three categories of supervised learning, unsupervised learning, and reinforcement learning, as well as important points about generative AI (copyright, confidential information, hallucinations).

Typical Past Question Patterns

• "Which of the following is the most appropriate example of supervised learning?" type
• "Which of the following is an appropriate precaution when using generative AI?" type

Related Terms

• IoT (What is IoT): A source for collecting data analyzed by AI
• Core technology for promoting DX (What is DX)

Study Tips

As a study tip, it is important to first visualize the inclusive relationship of AI ⊃ ML ⊃ DL. Learning methods can be categorized on two axes: whether there are correct answer labels, or whether they operate on rewards. Also, be sure to memorize the risks of generative AI (hallucinations, copyright, information leaks) as a set.

Summary

If you grasp the inclusive relationship, the three categories, and the risks of generative AI, you can handle nearly all frequently asked questions. For comprehensive practice on the Technology domain, see the Technology Summary, and for full-length practice, use the Practice Exam.