Understand artificial-intelligence systems

Definition

Understanding in principle how generative AI systems work, their biases, their limits and their effects on the creative value chain. This competency is the understanding side of AI; its productive use belongs to competency 4.3. You only direct well what you understand.

Components

• The statistical principle of generative models: predicting probable outputs, not producing truths
• Training data: provenance, embedded biases, blind spots
• Structural limits: hallucinations, sycophancy, sensitivity to phrasing
• Mapping capabilities: what these systems do well, badly, or in a deceptively plausible way
• Effects on creative occupations: the collapse of production cost, the displacement of value upstream (the eye) and downstream (the relationship)
• Systemic stakes: rights over training works, energy footprint, concentration of actors

The four levels

1. N1 Discovery Lower secondary · complete beginner

   Is able to explain that a generative AI produces probable outputs rather than truths, and to cite an error they have observed.

   Evidence of assessment

   Show and explain a hallucination or bias encountered in real use.

2. N2 Application Upper secondary · advanced beginner

   Is able to anticipate which types of task an AI handles reliably, riskily, or not at all.

   Evidence of assessment

   Build a personal grid of “tasks to delegate / tasks to verify / tasks to keep”, argued from real trials.

3. N3 Adaptation Tertiary · junior professional

   Is able to analyse the effects of an AI system on a given creative practice: quality, costs, lead times, rights.

   Evidence of assessment

   Carry out an impact study of introducing an AI tool into a real creative process.

4. N4 Orchestration Experienced professional

   Is able to inform collective choices regarding AI: adoption, usage charter, training needs.

   Evidence of assessment

   Produce an adoption recommendation or an AI usage charter actually applied by a team or an organisation.