What is GLM?
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.
Here, GLM refers to the AI model family offered by Z.ai. A model is the trained component that turns supplied information into responses or predictions. It names a model family, not a general agent workspace.
The abbreviation needs context because GLM is also used in statistics for generalized linear model, a family of mathematical methods for relating inputs to an outcome. That statistical usage is different from the named language-model family discussed here. Readers should not assume that the same letters identify the same kind of system.
Z.ai’s developer platform provides access to models through an API, an agreed interface for software requests and responses. The available models can support different tasks and kinds of information. Multimodal means working with more than one type, such as language and images. The exact support depends on the selected model.
For example, a developer might connect an application to a GLM model to draft descriptions from supplied product information. The model generates the wording. The application determines which records are supplied, where the draft is stored and whether anybody must approve publication. Those operating decisions do not come automatically from the model-family name.
Coding and agent tasks involve the same distinction. Code is the set of instructions a computer executes. An agent is software that selects actions while pursuing a goal. A model can help propose code or actions, while the surrounding environment supplies tools, records progress and controls access.
A provider’s description of a model’s strengths is useful for understanding intended use, but it is not a substitute for checking the task that matters to you. A benchmark—a standardized collection of comparison tasks—also measures only what its design supports.
Identify the model, the service running it and the application through which you work. GLM names one possible model family within that arrangement. To decide whether it is useful, test a specific version on the work you need to do.
Source: Z.ai — Developer Overview.
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