When Alan Turing was pondering artificial intelligence in the mid 20th century, he reached a crucial impasse on what it meant for a machine to be intelligent. It was the idea that a machine did not have to be perfect in order to be intelligent. Many thinkers at his time pondered the question through the lens of an infallible machine. Turing did not.
“Instead of trying to produce a programme to simulate the adult mind,” he asked, “why not rather try to produce a programme to simulate the child’s?” This meant that a complete and closed off machine was not the answer. Instead, a machine that could learn and adapt was the key.
Infallibility, in his words, “is not a requirement for intelligence…What we want is a machine that can learn from experience. The possibility of letting the machine alter its own instructions provides the mechanism for this.”
If you think about it, Turing’s mechanism exists in our everyday music experience. Most of us deal with a program like Pandora or Sonos. It requires that we enter in a suggestion (artist, band, etc.) to create a station. From there, song after song comes. But sometimes it throws in a sour selection. This is where the skip or thumbs down feature comes into play. When this is chosen it is a matter of reinforcement, letting the program know to avoid the song and others like it in the queue.
This won’t stop Pandora from choosing terrible songs however. We are dealing with a fallible machine here. But the ability for adaptation, to reinforce and dismiss choices, is what gives Pandora the mechanism for developing intricate musical patterns over time.
One has to wonder, beyond the noticeable objections, whether programs like Pandora have and develop musical taste like we do: through learning from experience what we enjoy.