How to check AI output: Human, AI and independent adversarial checks

In this class

What this class covers

1. 01Request revision histories and source evidence to make document outputs easier to verify
2. 02Anticipate potential errors and define corresponding verification criteria
3. 03Check document outputs using quantities, reverse checks, and differences
4. 04Choose programmatic or AI checks based on whether a definitive answer exists
5. 05Learn to build an AI verifier independent of the executor through a supervisory agent example

Instructor

Learn with Peggy Lo

Peggy Lo

Peggy Lo

Administrative Workflow AI Practitioner

• Moved from a non-technical background into AI coding
• Frontline experience across government, NGOs, caregiving, and parenting
• Explores how AI agents can help everyday workers and people with disabilities solve real problems

Workflow · 5 classes

5 courses in Workflow

One price covers all 5 classes. Start with the one you want.

1. 01 AI's two limits: Too much to read, different every timeEric Wu · About 30 min →
2. 02 How to check AI output: Human, AI and independent adversarial checksPeggy Lo · About 25 min This course
3. 03 Working with AI like an engineer: Specs, task breakdown and controlled deliveryRichard Lee · About 20 min →
4. 04 AI cost control in practice: Tokens, models and run budgetsWisely Chen · About 30 min →
5. 05 Sort workflow steps: Break down a customer-support flowEric Wu · About 40 min →