REUTERS/Kim Kyung-Hoon
SoftBank humanoid robot known as Pepper dressed as a waiter moves its hand at Pepper World 2016 Summer during SoftBank World 2016 conference in Tokyo, Japan, July 21, 2016
Jonathan Rochelle, head of the product management team for Google's education outreach arm, called Google for Education, says the techniques we use to teach our machines are far more effective at making them smarter than the methods we use to teach living, breathing students.
Whereas machines learn algorithms and systems for building knowledge, kids get facts and equations jammed down their throats, he says.
"We're not teaching them how to learn," Rochelle tells Business Insider.
Much of the science about learning suggests that Rochelle's criticism is well-founded. Cognitive science has shown that in order to grow into capable adults, kids shouldn't just learn when George Washington crossed the Delaware or how to compute the area under a curve. They also need to develop the skills that let them synthesize that raw information and draw conclusions when given a specific problem.
As Rochelle explains, that's basically how computer algorithms work. IBM Watson is so smart because it continuously takes in new information and fits it into the frameworks it was given at birth. When it encounters a foreign query, it can make an educated guess based on all that prior knowledge.
The good news for our kids is that teaching methods - like programming strategies - can change. Programming didn't always take the algorithmic approach, Rochelle says. Thirty years ago, the field of machine learning looked a lot like how pedagogy works today. Engineers gave their rudimentary machines explicit instructions. They told them what to know and provided few frameworks for them to make sense of new inputs on their own.
But then developers got smarter about teaching their machines - they experienced "machine-learning learning," as Rochelle puts it.
Educators could do well to adopt the same principles. In practice, that means helping kids understand why a particular math equation works and when it's applicable, rather than just teaching the formula.
When students are learning that way, Rochell explains, "you don't have to feed them the actual answers. ... It's more, here's information that got us to a prior conclusion. Now we're going to teach you how to draw conclusions."