Lesson 7 of 10
Generative AI and large language models
Generative models create content by learning statistical patterns from training data. A large language model (LLM) processes text as tokens and generates a continuation based on context. It can help draft, summarize, classify, or transform text, but plausible-sounding output can still be incorrect, incomplete, or outdated.
Use an LLM for tasks where generated text can be reviewed. Do not treat fluent output as evidence, and do not paste confidential or personal information into a service unless its data handling has been approved. Verify names, figures, citations, and important claims against trusted sources.
Match model to task
For a fixed label set, a small classifier may be cheaper and more consistent. For open-ended summaries, a generative model may be useful. Consider quality, latency, privacy, cost, and the availability of human review.
Try it: Write a one-sentence task where generation is appropriate and another where an ordinary deterministic program is safer.