What is GenAI?

Understanding generative AI, how it works, and its implications for higher education

Generative AI (GenAI) refers to a category of AI systems that generate new content, such as text, images, or audio, in response to user input. This is different from non-generative tools such as Grammarly or traditional spelling and grammar checkers, which mainly analyze and improve existing text instead of creating new content.

This technology is evolving rapidly, and its outputs are increasingly difficult to distinguish from human work, which raises understandable questions in higher education. When used responsibly, GenAI can support teachers' workflows and help generate ideas for learning activities or lesson design, and it can act as a practice or feedback partner for students.

To support responsible use, UvA has developed its own AI Chat for academic purposes, offering a safer and more reliable alternative to commercial tools such as ChatGPT or Claude. With clear guidance from teachers and thoughtful course design, GenAI can contribute positively to learning and the development of the skills students need.

A Conceptual Model of AI Domains

AI is often discussed as a single concept, but different types of AI do different kinds of work. Generative AI (GenAI) is a specific subset of AI, alongside machine-learning systems that analyze data to classify, score, or predict outcomes.

Machine-learning systems typically produce constrained outputs, such as probabilities or categories. For example, a streaming platform recommending a new series based on what you have watched and liked before.

GenAI, by contrast, is designed to generate new content, such as text or images, in response to a prompt. Tools like ChatGPT are built on large language models, which produce fluent and plausible outputs but do not evaluate correctness or understanding.

Why this distinction matters: It affects what AI should appropriately be used for, including when its use supports learning goals and when it risks bypassing them. Where machine learning can help identify patterns or risks in existing data, GenAI is better used to support explanations, practice, feedback or idea generation – when paired with intentional educational design.

How GenAI and LLMs Work

GenAI that generates text (e.g., ChatGPT, Claude, etc.) is based on Large Language Models (LLMs) which learn language patterns from 'training data' – a vast collection of existing texts in books, articles, and websites.

Through probability, LLMs predict which word is most likely to come next in a sentence, and then sample from these possibilities, rather than always choosing the single most likely word.

This process enables LLMs to generate complex and well-written texts (called 'output') in different styles and languages in seconds, based on a user prompt (set of instructions or a question).

GenAI Limitations

Despite its many benefits, GenAI has some problems. Users should be mindful of functional issues and ethical concerns when using this technology, and they should use critical thinking to analyze and evaluate the output generated by GenAI.

  • •Hallucinations: GenAI can generate false information that sounds convincing
  • •Biases: Training data can contain and reproduce societal biases
  • •Privacy concerns: Commercial tools may store and use your data in unclear ways
  • •Sustainability: AI systems consume significant energy and computational resources

Next Steps

Once you have a deeper understanding of what GenAI is and how it works, you can begin to plan how you might effectively design for or around its use, and build your AI literacy.

Build Your AI Literacy