A new book, by the inventor of the Palm Pilot, uses music to unravel the mysteries of human intelligence. You'll hear about it in this Science Update.
The music of the mind. I'm Bob Hirshon and this is Science Update.
(Music clip: Beethoven's 5th, pitch-shifted to E flat minor)
You're listening to Beethoven's Fifth Symphony—in the wrong key. You probably recognized it anyway. That's because your brain files Beethoven's Fifth as a pattern, not a list of pure notes.
In his new book, The World Is a Song, neuroscientist Jeff Hawkins uses our musical savvy to model all human intelligence.
The way we learn music is no different than the way we learn language, and it's no different than the way we learn to move our body in the world, and it's no different than the way we learn to recognize the things we see.
He describes a brain that relentlessly looks for patterns over time, like melodies; creates hierarchical structures, like verses and choruses; and makes predictions based on these stored patterns, just as we anticipate the next note of a tune.
Next on Science Update, we'll tell you how this could change technology. I'm Bob Hirshon for AAAS, the Science Society.
Making Sense of the Research
Unless you have perfect pitch, you probably didn't notice there was anything unusual about the music at the beginning of this story. Yet, none of the notes you heard were the correct notes for Beethoven's Fifth. How did your brain recognize it? By listening to the duration and rhythm of the notes, and to the relative pitches (whether, and by how much, each note is lower or higher than the last). In other words, you remember "Da-da-da-dummmm" rather than G-G-G-E flat (the actual notes). Because you store this memory as a pattern, it's much more flexible: you can recognize the song even if it's played in a different key, or by different instruments, or even hummed badly by your dad.
This all might sound unremarkable, but it's really a tremendous achievement. Imagine trying to program a computer to recognize a melody, or a room, or a chair, or the sound of a dog barking—all without being able to write hard and fast rules about them. It's a very complex problem and scientists have long struggled to understand how the brain does it.
Hawkins uses music as a model for a couple of reasons. First, it's something that everyone can intuitively understand and relate to. And second, music is a pure pattern. It has no physical component (unlike, for example, a dog barking, which is linked to the physical being of a dog). It has no smell, no taste, nor any tactile quality. Music itself has no linguistic meaning (although song lyrics do). It's just a sequence of tones. And that makes it relatively easy to study.
Imagine hearing a piece of music for the first time. Research has shown that it takes a while to figure out the patterns: when a phrase or movement starts and ends, which is the verse and which is the chorus. But immediately, your brain starts making comparisons. "This is like that other Mozart piece I've heard," or "This sounds like that Beastie Boys song." Based on these comparisons, you start anticipating where the next note of the piece will go. All the while, your brain records whether your guesses were correct or not and revises its predictions based on that new information.
The next time you hear the song, it starts to make more sense as a whole. You remember more specific details and recognize small patterns like guitar licks and drum loops within the larger pattern of the song itself. These patterns fit into a hierarchy, like nesting dolls, little structures within big structures, until you know the song inside and out.
Hawkins argues that we learn anything else in much the same way; it's just harder to study. Think of a time you've visited a strange city. You might start by comparing it to other cities you've visited by recognizing how it's similar and how it's different. You might start by learning to recognize isolated streets without understanding how they fit together. As you get more familiar with the city, you start to build a mental map, and creating a hierarchical memory bank of neighborhoods, blocks, streets, and landmarks.
By using music as a model, Hawkins hopes to learn more specifics about how this process works in the brain. Then, inventors can use this information to build smarter computers that think in flexible ways like people do.
Now try and answer these questions:
- Why is music a good model for intelligence?
- Describe the process of learning a new piece of music, using the ideas of hierarchical patterns and making predictions based on stored patterns.
- Compare learning a piece of music to learning something else not described above.
- How does human learning differ from simple machine learning? Why would it be difficult to teach a computer to recognize a popular song, like "Somewhere Over the Rainbow?"
The MIND Institute is devoted to studying the relationship between music and intelligence and using music to prepare K-12 students for success in school, work, and society.
“The World Is a Song” is taken from Hawkins’ book On Intelligence with Sandra Blakeslee.