1 · What AI really is (and isn't)

Today's popular AI (chat assistants like ChatGPT, Claude, and Gemini) are large language models (LLMs). At their core they predict the next most-likely chunk of text based on patterns learned from huge amounts of writing. They are extraordinary pattern-matchers, not knowers of truth.

That one fact explains almost everything you need to be careful about:

• They sound fluent and confident even when they are wrong: fluency is what they optimize for, not accuracy.
• They are strong at drafting, summarizing, rephrasing, brainstorming, translating, and explaining.
• They are weak at exact facts, fresh/current events, math, counting, and anything needing a verified source, unless connected to tools that look things up.

Modern LLMs are built on the transformer architecture (Vaswani et al., 2017). They don't "look things up" in a database by default; they generate the most plausible-sounding answer. Plausible is not the same as correct.

Mindset for this course: AI is a powerful assistant, not an oracle. You stay the editor, the fact-checker, and the person responsible for the result.

Sources

• Vaswani, A., et al. (2017). Attention is all you need. arXiv:1706.03762. https://arxiv.org/abs/1706.03762
• National Institute of Standards and Technology. (2023). AI Risk Management Framework (AI RMF 1.0). https://www.nist.gov/itl/ai-risk-management-framework