Explore the concept of teaching machine learning systems in this thought-provoking conference talk by Jussi Karlgren. Delve into the current hands-off approach to AI learning and its limitations, examining why we often resort to retraining and data improvement when AI systems behave unexpectedly. Contrast this with the potential benefits of feature engineering and instruction in machine learning. Gain insights from Karlgren's extensive experience in language technology and information retrieval research, and consider how a more guided approach could enhance the quality and reliability of AI outputs. Recorded at the 2023 GAIA Conference, this 28-minute talk challenges conventional wisdom and offers a fresh perspective on improving machine learning systems.