Xiao-Li Meng and Liberty Vittert at the Harvard Data Science Review:
Xiao-Li Meng: Well, thank you so much, Steve, for joining us. I know how busy you are, so let’s just get to it. What is intelligence? What are the key components? What do you think constitutes intelligence?
Steven Pinker: I think intelligence is the ability to use knowledge to attain goals. That is, we tend to attribute intelligence to a system when it can do multiple things, multiple steps or alternative pathways to achieving the same outcome: what it wants. I’m sitting here right now in William James Hall, and my favorite characterization comes from William James himself, the namesake of my building, where he said, “‘You look at Romeo pursuing Juliet, and you look at a bunch of iron filings pursuing a magnet, you might say, ‘Oh, same thing.’ There’s a big difference. Namely, if you put a card between the magnet and filings, then the filings stick to the card; if you put a wall between Romeo and Juliet, they don’t have their lips idiotically attached to opposite sides of the wall.” Romeo will find a way of jumping over the wall or around the wall or knocking down the wall in order to touch Juliet’s lips.’ So, with a nonintelligence system, like physical objects, the path is fixed and whether it reaches some destination is just accidental or coincidental. With an intelligent agent, the goal is fixed and the path can be modified indefinitely. That’s my favorite characterization of intelligence.
More here.