AI Foundations - lesson 3

What it gets wrong, and how to catch it

Confidence is not evidence. Here is where these systems reliably fail.

What you will take away:

• Fabrication is a feature of how these systems work, not a bug that will be patched away.
• The highest risk outputs are the ones that sound most authoritative: names, numbers, dates, citations, legal and medical claims.
• Bias in output reflects bias in training data, which underrepresents most African contexts.
• A thirty second verification habit removes the large majority of real world risk.

Afrilens - global AI with an Africa lens