What AI really is, from model to agent, opportunities and risks, adoption in a company, data culture, controlled experiments and cloud.
- 1
What Artificial Intelligence Really Is
Artificial intelligence, machine learning and deep learning are not synonyms: Mitchell's definition, empirical risk, overfitting measured with code, and the three learning paradigms.
23 min - 2
From Model to Agent
The perceive-decide-act loop, the five agent types, a minimal Python agent with tools and a step limit, and the difference between a workflow and an agent.
19 min - 3
AI Opportunities, Challenges and Risks
Accuracy that lies, auditing a model by group, the proven incompatibility of fairness criteria, and an essential map of GDPR, the AI Act and the NIST AI RMF.
21 min - 4
Implementing AI in a Company
From problem to decision with costs in hand: the five-field card, the CRISP-DM cycle, the optimal threshold derived from costs and checked in code, and the build, buy or skip choice.
21 min - 5
Data Culture
How to turn a vague problem into a question data can answer: outcome, driver and guardrail metrics, Simpson's paradox and Goodhart's law with exact numbers.
21 min - 6
Test and Learn: Controlled Experiments and A/B Tests
Randomisation, A/B tests, p-values, the sample size formula checked by simulation, and the peeking trap shown with a repeated experiment.
22 min - 7
Cloud and Artificial Intelligence
What the cloud is according to the NIST definition, IaaS, PaaS and SaaS with their split of responsibility, and elasticity and the buy-versus-rent break-even computed in Python.
20 min