Machine Learning - What's Surprising About That?

This is the second part of the episode about Robert Haskell's book Transfer of Learning.

In this part, we go in detail into the importance of rich declarative knowledge for transfer. Topics include:

• The difference between surface structure and deep structure of problems, and how experts can see through the former to get at the latter, allowing transfer to happen
• How pure scientific discoveries with no application resulted in groundbreaking technological breakthroughs decades later
• Prolific inventors' advice on what it takes to be a great inventor (hint: "irrelevant" knowledge)
• How even specialists in procedural knowledge ("skills") openly admit that rich, well-structured declarative knowledge is the cornerstone of transfer

Enjoy the episode.