Generative Artificial Intelligence

"Generative AI (noun): artificial intelligence... that is capable of generating new content (such as images or text) in response to a submitted prompt (such as a query) by learning from a large reference database of examples"

- Miriam Webster

In recent years, generative AI has grown rapidly from a subfield of artificial intelligence to a core paradigm for tools disrupting how we learn, teach, and work in higher education. Note that Merriam Webster's definition includes the following components:

1. Generative AI generates new content - everything from text in chat interactions to images, music, video, documents, datasets, and code
2. Generative AI responds to a prompt - a specific request from a user, usually written as text (or verbalized)
3. Generative AI learns from a large reference database of examples  - this includes a large corpus of texts like websites and books, video content, and other artifacts, especially content scraped from the open Internet. This training is supplemented by annotation and feedback testing done by humans.

You can also think of generative AI as a predictive model - that is, a specific, prepared model that makes predictions of the most desirable output given a user's prompt. 

(Note in this guide, we use generative AI and genAI interchangeably. Some outlets use different / idiosyncratic stylizations - e.g. the New York Times stylize the term as generative A.I.)

Generative AI is the subject of widespread societal, technological, and commercial focus in 2025. But it comes from a longer history of artifiicial intelligence that spans decades.

The following visualization represents that historical and conceptual context:

Some limitations and causes for concern:

• With some exceptions, we don't have a lot of information about how the models that drive these tools were trained. That means that, in most cases, users have no way of knowing what information an AI tool has had access to. 
• Most AI tools capture and reuse a significant amount of user data. You should not share private information with chatbots or similar AI tools.
• All AI text generators can, and often do, produce plausible sounding but false information (known as "hallucinations"), and by their nature will produce output that is culturally and politically biased.
• There are a number of ethical and societal concerns about generative AI, from the way these tools were created, to the way they are being applied now and in the future.

Amherst College Library identifies six key categories of social harms and costs:

1. bias, misrepresentation, and marginalization
2. labor exploitation and worker harms
3. misinformation and disinformation
4. privacy violations and data extraction
5. copyright and authorship issues
6. environmental costs

Read the complete annotated breakdown with sources here