AWS Certified AI Practitioner AIF-C01 Practice Question

Foundation models are very large neural networks pretrained on extensive and varied data so they develop general linguistic or visual capabilities. After pretraining, the same model can be adapted with comparatively small amounts of additional data to support many downstream tasks, including summarization and translation. Embeddings are vector representations, diffusion models specialize in iterative data generation (often for images), and a vector search index is a retrieval structure; none of these alone provide the broad, adaptable capabilities described.