Prepare and test your skills
Prepare and test your skills
Free sample question, answered and explained below. The Generative AI Leader practice bank has 2,191 more like it. Try 10 free: instant score, no account.
An enterprise is planning to deploy a domain-specific generative AI application on Google Cloud to handle specialized customer support workflows. The project team is assessing data acquisition costs and preparation efforts across different training strategies.
How do generative AI foundation models leverage unlabeled and labeled data across their lifecycle to optimize data collection resources?
This option claims that generative AI models operate in an entirely unsupervised manner across all development phases, removing the need for labeled data entirely.
While prompt engineering can bypass fine-tuning, when fine-tuning is performed to specialize model behavior, labeled datasets are necessary.
This option suggests an inverted training pipeline where foundational pre-training relies entirely on human-labeled data and downstream adaptation requires only unlabeled data.
Pre-training leverages self-supervised learning on vast unlabeled data, which is what makes large-scale foundational AI scalable.
This option asserts that foundational pre-training depends on a 1:1 balance of labeled and unlabeled datasets simultaneously to establish core language understanding.
Understanding the sequential shift from massive unlabeled data in pre-training to compact, curated labeled data in fine-tuning is fundamental to evaluating AI resource implications.
Generative AI foundation models are built using a two-stage approach: extensive unsupervised pre-training on massive text or multimodal datasets, followed by supervised fine-tuning (or model tuning) on targeted, high-quality data examples.

Answer them yourself and get your score the moment you finish. No account needed.
Full access includes
Keep the momentum going with these hand-picked practice scenarios
Last updated