The key topics in artificial intelligence can be read from the Association for Computing Machinery (ACM) classification system, which splits the field into nine top-level areas: natural language processing, knowledge representation and reasoning, planning and scheduling, search methodologies, control methods, philosophical and theoretical foundations, distributed AI, computer vision, and machine learning. The companion site of the textbook Artificial Intelligence: A Modern Approach reproduces the 2012 version of that list and ties each area to chapters of the book.
Key takeaways
- The 2012 ACM classification, as shown on the textbook's site, has nine top-level AI areas, and machine learning has by far the longest list of subtopics.
- Classic areas such as search, planning, and reasoning still sit beside machine learning in the same taxonomy.
- Intelligent agents are filed under distributed AI, next to multi-agent systems, and map to Chapter 2 of the textbook.
- A real product usually draws on several areas at once, so labeling it with a single topic is a simplification.
- The list is a way to organize research, not a ranking of importance or a forecast of where the field is going.
The nine areas and where the textbook covers them
The page AI topics in the ACM Computing Classification System on the Berkeley-hosted site of Stuart Russell and Peter Norvig's textbook states that it presents the categories and subcategories of AI topics from the 2012 ACM scheme. For each entry it gives the chapter or section of the book where the topic is covered. The table condenses it, with a few subtopics chosen from each area.
| Area | Example subtopics listed | Textbook chapters |
|---|---|---|
| Natural language processing | Machine translation, speech recognition, information extraction | 23 and 24 |
| Knowledge representation and reasoning | Description logics, probabilistic reasoning, ontology engineering | 10, with 13 to 15 for probability |
| Planning and scheduling | Planning for deterministic actions, planning under uncertainty | 11 |
| Search methodologies | Heuristic function construction, game tree search, randomized search | 3 to 5 |
| Control methods | Motion path planning, computational control theory | 26 |
| Philosophical and theoretical foundations | Cognitive science, theory of mind | 27 |
| Distributed artificial intelligence | Multi-agent systems, intelligent agents, cooperation and coordination | 18, with 2 for agents |
| Computer vision | Image and video acquisition, computer vision tasks | 25 |
| Machine learning | Supervised, unsupervised, and reinforcement learning; neural networks | 19 to 22 |
Why machine learning looks bigger than the rest
On the textbook's list, machine learning is subdivided into supervised learning, unsupervised learning, reinforcement learning, multi-task learning, learning settings, learning approaches, and learning algorithms, each with its own subtopics. Examples under those headings include classification and regression trees, support vector machines, neural networks, Q-learning, and cross-validation. The other eight areas have between two and a dozen subtopics.
That imbalance reflects how the taxonomy was drawn, not a measure of how much work each area receives. A classification scheme gives fine detail where its authors expected many papers to need labels. It also dates from 2012, so some terms used today, such as large language model, do not appear as headings. Text generation of that kind falls, under this scheme, somewhere between natural language processing and neural networks within machine learning.
Sorting familiar systems into the areas
Most products draw on more than one area. The table below is an illustrative classification made by this article, not part of the ACM scheme. It shows how a single label hides the mix.
| System | Main areas involved | Why |
|---|---|---|
| Email spam filter | Machine learning | Learns from labeled examples to classify messages |
| Map route finder | Search methodologies | Looks for a low-cost path through a graph of roads |
| Chess engine | Search methodologies, machine learning | Game tree search, often with a learned position evaluation |
| Voice assistant | Natural language processing, machine learning | Speech recognition feeds a language component that picks a response |
| Warehouse robot | Computer vision, planning, control methods | Perceives its surroundings, plans a route, and drives the motors |
What the map says about implications
Each area carries its own questions about reliability and risk. Machine learning raises the matter of training data and error rates across groups. Natural language processing raises the question of fabricated statements. Computer vision raises questions about identification and surveillance. Knowledge representation and planning, whose rules can be read by a person, make it easier to audit why a system acted as it did. Distributed AI raises the question of how several agents behave together, which a single-agent test does not reveal.
These links are interpretive. The taxonomy lists topics and does not discuss harms. For specific risks, such as hallucinated output or copyright disputes, the better evidence comes from case records, regulators, and studies rather than from a topic list. The related articles below cover several of them.
A short way to use the map
- When a product is described as AI, ask which of the nine areas it relies on and which methods inside them.
- When a claim is made about "AI" in general, check whether the evidence comes from one area only, such as image classification.
- When reading a course description, compare its headings with the list above to see what is missing.
FAQ
What are the main topics of artificial intelligence?
Under the 2012 ACM scheme as reproduced on the textbook site, they are natural language processing, knowledge representation and reasoning, planning and scheduling, search methodologies, control methods, philosophical and theoretical foundations, distributed AI, computer vision, and machine learning.
Is machine learning the same as AI?
No. In this taxonomy machine learning is one of nine areas. Search, planning, and logical reasoning are separate areas that do not depend on learning from data.
Where do chatbots fit?
Chatbots belong mainly to natural language processing and use machine learning, since the 2012 scheme predates the current generation of language models and has no separate heading for them.
Is the ACM list the only way to divide AI?
No. University curricula, conference tracks, and textbooks each group topics differently. The ACM list is one widely used reference point, and others overlap with it without matching exactly.
Related reading
- Contested questions in AI ethics
- Why AI systems state false things
- The impact of AI on modern society
- AI and copyright law
More in AI Basics.
Last verified: October 2026
Sources
- Categories and subcategories of AI topics from the 2012 ACM Computing Classification System, Artificial Intelligence: A Modern Approach companion site, University of California, Berkeley
- Russell S, Norvig P, Artificial Intelligence: A Modern Approach, 4th US edition, companion site, University of California, Berkeley, 2022 page update