Cheat sheetGCP-01

Gen AI & LLM Fundamentals

GCP GenAI Leader / Gen AI & LLM Fundamentals

The shared vocabulary of the whole stack: generative AI makes new content, LLMs are adaptable foundation models, and everything is measured in tokens.

Generative vs traditional
What is itGenerative AI creates new content; traditional ML classifies or predicts. Pick the right tool for the goal.
Tokens
Unit of everythingText is split into tokens; cost and context limits are counted in tokens, not words.
Embeddings
Meaning as vectorsVectors that place similar meanings close together, powering semantic search and retrieval.
Hallucination
Core riskFluent output can be false or unsupported; this motivates grounding and human oversight.
⚑

Before proposing an LLM, ask whether the task truly needs generated content or whether classic ML or deterministic code is safer and cheaper.

Good fitSummarize thousands of reviews into readable digests -- generated language at scale.
Poor fitCompute an exact invoice total -- keep this in deterministic code, not the LLM.
Generative AI = new content; traditional ML = label or number.
Tokens drive cost and context limits; one word can be several tokens.
Embeddings = meaning vectors; hallucination = confident but wrong output.
fundamentalsllmtokensembeddingshallucination
review in 6d