by James Bailey |

Stephen Wolfram is already helping us appreciate the possibilities of bringing the latest AI technology deep into the K-12 curriculum. In his blog entry Machine Learning for Middle Schoolers, he shows how preteens can do some amazing things with the building blocks already in the Wolfram Language. Here we will take a slightly different approach, based on the belief that the real benefit comes by enabling kids to understand what is under the hood, and then to write their own little learning algorithms and watch them do their thing. The payoff spans the curriculum. (If the liberal arts are not about learning, they are not about much of anything.)

Can kids really get over the hurdles of advanced mathematics and build an app that learns? Potentially they can because they no longer have to grind it out all by themselves with pencil and paper as their grandparents did. Think of the computer as reaching out and taking control of the student's pencil during the rough spots. Once it has steered around the pot holes, it lets go and encourages the student to press ahead on their own.

Prof. Andrew Ng, a cofounder of Coursera, recently launched a four-month online course in Deep Learning. He sees it is a very big deal, the kind of big deal every student needs to know about because it will pervade their lives:

"

AI is the new electricity... Over the next decade I think all of us have an opportunity to build an amazing world—an amazing society—that is AI powered. I hope that you will play a big role in the creation of this AI-powered society.

"

Already in its third week, his course has us ginning up simple neural networks and using them to find cats in images. Prof. Ng has worked hard to eliminate many of the mathematical obscurities that make AI hard for the rest of us. But he hasn't eliminated them all. As he confesses:

"

In case it seems like I'm getting [into] a lot of nitty gritty linear algebra, it turns out that being able to implement this correctly is important in the Deep Learning era.

"

He is being too much of a fuddy-duddy here. Current AI practitioners, including Prof. Ng, came from the ranks of mathematics concentrators. They had to sweat all kinds of arcane mathematics in their own student days and seem to believe it was good for them. In their eyes, those of us that follow must (damn well) do the same. Prof. Ng is actually much freer of these hangups than most, certainly freer than the Calvinist practitioner he interviews in Week Three:

"

Linear algebra and probability are very important… You've got to master the basic math that underlies the whole approach in the first place [even if] it might be a bit of a painful experience.

"

No you don't. Linear algebra and probability are exactly the places where a self-driving computer needs to gently take hold of the pencil and steer for a while, thereby letting a wide range of students experience Deep Learning technology firsthand. And we’re gonna need that wide range, as this practitioner admits:

"

Do you want to make sure that the societal issues surrounding AI work out well, that we’re able to make sure AI benefits everyone rather than causing social upheaval and trouble with loss of jobs?

"

If there's one lesson we've learned from global trade of late, it's that societal changes that are understood only by a few never end up benefiting everyone. Wide benefit follows wide understanding.

So come along for the next few months as we engage this course week by week. What is the course teaching that every student needs to know? What are the mathematical hurdles that prevent the course as it now stands from becoming integral to the K-12 curriculum? What will it take to eliminate those obstacles? Where are the Silicon Valley “school reformers” that would like to really make a difference in our schools?

tags:

Comments