What parallels exist between AI programming and pedagogy?
Today, I had to teach my part-timer how to delete books from our inventory. This is a two-phase process: delete the book from our inventory records then delete the book from our interlibrary loan records. My PTer is an older woman not at all versed in computers, so to teach her, I first demonstrated the necessary steps, then asked her to do it while I guided her, then asked her to do it alone. She understood the central steps and began to delete books at a reasonable rate.
A few minutes in, she hit the back button one too many times and came upon a screen that was unfamiliar to her. The screen had buttons leading back to the interface she needed to use. They were very clearly labeled. But she could not understand the information in the labels, either because she had shut down all “receiving” from the direction of the screen in a panic or because she did not want to try for fear of “messing the computer up.”
Helping her with this made me think of the problems AI programmers have. They cannot tear levers from their mind and give that set of inferences to an AI wholesale. They cannot say “the AI will KNOW that, if it hits back once too many times, to just hit the button that says ‘Delete Holdings.’ After all, its job is to delete holdings so it knows that the ‘Delete Holdings’ interface is the one it needs.” Just like my PTers, in order to make that inference, the AI must be able to receive information about this new surrounding, process that information, and infer from it how to obtain its goal (i.e. getting back to ‘Delete Holdings’).
What sort of lessons and parallels could be drawn from AI programming that would be useful in pedagogy? I will admit I am ignorant of AI theory or practice save what I have picked up from the Sequences. But, the overlaps seems worth exploring. Indeed, I suspect others have explored it before me. The Sequences are certainly didactic. I also wonder if teaching (especially teaching those who are technologically illiterate) would be a useful experience to those planning to work in AI programming and ethics.
I keep thinking about the relations between machine learning and human learning, esp. teaching children a lot. Basically all results in one field carry over to some degree to the other. Some things only apply on the neuronal level. Others only in very specific settings.
What parallels exist between AI programming and pedagogy?
Today, I had to teach my part-timer how to delete books from our inventory. This is a two-phase process: delete the book from our inventory records then delete the book from our interlibrary loan records. My PTer is an older woman not at all versed in computers, so to teach her, I first demonstrated the necessary steps, then asked her to do it while I guided her, then asked her to do it alone. She understood the central steps and began to delete books at a reasonable rate.
A few minutes in, she hit the back button one too many times and came upon a screen that was unfamiliar to her. The screen had buttons leading back to the interface she needed to use. They were very clearly labeled. But she could not understand the information in the labels, either because she had shut down all “receiving” from the direction of the screen in a panic or because she did not want to try for fear of “messing the computer up.”
Helping her with this made me think of the problems AI programmers have. They cannot tear levers from their mind and give that set of inferences to an AI wholesale. They cannot say “the AI will KNOW that, if it hits back once too many times, to just hit the button that says ‘Delete Holdings.’ After all, its job is to delete holdings so it knows that the ‘Delete Holdings’ interface is the one it needs.” Just like my PTers, in order to make that inference, the AI must be able to receive information about this new surrounding, process that information, and infer from it how to obtain its goal (i.e. getting back to ‘Delete Holdings’).
What sort of lessons and parallels could be drawn from AI programming that would be useful in pedagogy? I will admit I am ignorant of AI theory or practice save what I have picked up from the Sequences. But, the overlaps seems worth exploring. Indeed, I suspect others have explored it before me. The Sequences are certainly didactic. I also wonder if teaching (especially teaching those who are technologically illiterate) would be a useful experience to those planning to work in AI programming and ethics.
I keep thinking about the relations between machine learning and human learning, esp. teaching children a lot. Basically all results in one field carry over to some degree to the other. Some things only apply on the neuronal level. Others only in very specific settings.
Some random pages to follow: http://lesswrong.com/lw/jol/rethinking_education/ Vygotsky’s http://en.wikipedia.org/wiki/Zone_of_proximal_development