What is artificial intelligence?
The Harvard Chan AI Glossary
In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.
Artificial intelligence, or AI, is the broad field of building computer systems that perform tasks associated with intelligence. These include recognizing patterns, interpreting language, solving problems and selecting actions. AI is not one product or one particular way of programming. It is a family of approaches to making computers perform such tasks.
An email filter, a system that recognizes spoken words and a program that plans moves in a game can all fall within that family. Their abilities may be very different. A system that excels at one of these tasks does not automatically possess the others, and calling it intelligent does not establish awareness or human-like understanding.
Some AI systems follow carefully designed rules. Others use machine learning, which means learning useful patterns from examples rather than having every possible rule written out by a person. The learned component is called a model. A model might learn which patterns of pixels—the tiny colored elements of a digital image—usually correspond to handwritten numbers.
Generative AI is one part of this field. It produces content such as text, pictures or computer code. Other systems primarily assign categories or estimate values. For example, deciding whether a message is unwanted and writing a reply to that message are different tasks, even when the same application offers both.
This distinction helps explain why not all automation is usefully described as AI. A timer that turns on a light at a fixed hour follows a predetermined instruction. A system that interprets an unfamiliar spoken request must handle much greater variation. The boundary is not perfectly sharp, and the label alone rarely explains how a product works.
Ask not only what a model can do, but what the complete system helps someone accomplish. A model may recognize a request correctly while the surrounding application sends it to the wrong place. The useful unit of judgment is therefore often the whole task, not an isolated answer.
A public dental-care pathway dashboard provides a useful check on the label. Displaying who has attended, begun treatment or returned for follow-up may mainly require reliable records and ordinary software. AI could help interpret free-text notes or explain patterns, but the dashboard need not use a model for every calculation. Begin with the care problem and the work required to address it. “Uses AI” does not tell us whether more patients complete the care they need.
If you would like to find out more about AI and improve your leadership skills, consider the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.
AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.
For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.
Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy