Entrepreneurs

Can AI fix education? We asked Bill Gates

The rise of smartphones has transformed the way students communicate and entertain themselves. But the classrooms they spend so much of their time in remain stubbornly resistant to transformation. On one hand, technology has long had a home in classrooms — I learned to type on an Apple IIe in the late 1980s.

But for most schools, the approach to teaching remains stubbornly one-size-fits-all: a single teacher delivering the same message to a group of about 30 students, regardless of their individual progress.

More from The Verge:
Why Prince was an unparalleled musical genius
The state of the gadget union
This is how material design looks on Chrome OS, rolling out now

Bill Gates is working to change all that. Through the Bill and Melinda Gates Foundation, Microsoft's co-founder and chairman has invested more than $120 million to date in a developing field known as "personalized learning." It's a diffuse set of initiatives, led mostly by private companies, to develop software that creates individual lesson plans for students based on their performance, coaching them through trouble spots until they have mastered the subject at hand. Teachers still play a central role in the classroom, but they do less lecturing and more one-on-one coaching.




Bill Gates
Kelvin Ma | Bloomberg | Getty Images
Bill Gates

The effort is led by a dizzying array of startups with terrible names — think "Learnosity" — but big companies are starting to pay attention. In 2014 Google launched Classroom, which lets teachers post class announcements, assign work to students, and collect and grade their assignments. And last year Facebook announced a partnership with Summit Public Schools, in which the Gates Foundation is an investor, to create personalized learning software and make it freely available.

This week Gates spoke at the ASU GSV Summit, an education technology conference in San Diego. In a standing-room-only speech, he laid out the foundation's vision for accelerating the adoption of personalized learning around the world. Gates asked investors to take a longer view in education than other fields, because of epic school district purchasing cycles. He asked school districts to speed up those cycles by using more pilot programs, and by supporting data standards that make it easy to compare the efficacy of different products. And Gates told entrepreneurs to invest in research around the efficacy of their products, producing data that will encourage other schools to adopt personalized approaches.

Last year I wrote about Facebook's efforts around personalized learning, and afterward Gates' people invited me to speak with him about his evolving thinking about education. A few hours after his speech, Gates bounded into a hotel room on the 38th floor of the Manchester Grand Hyatt and sat in a chair by the window. We were joined by a Gates Foundation spokeswoman and, on the other side of the window, a lone seagull who observed our interview with great interest.

"It's still early stages," Gates said about personalized learning. "In five years, 10 years from now, will it be highly penetrated? That's not absolutely clear."

This interview has been edited and condensed.

Casey Newton: I think it's fair to say that even people who love tech don't always pay close attention to the ways it's transforming education. So at a high level, what is personalized learning doing for students at the schools where it's being tried? And what opportunities do you see it creating over time?

Bill Gates: Well the term "personalized learning" doesn't have an exact definition. In general, the idea is that people progress at a different rate. If you're ahead of what's being taught in the class, that's not good, you get bored. If you're behind, then they're using terms and concepts that create a general impression of "Hey, I'm not good at this." And science and math in particular — if they're talking about something you haven't had the explanation on, you just really give up in that area. And there is no way that you are brought back into it.

For me, one of the most interesting areas you're focusing on is remedial education in community colleges. Kids are asked to spend lots of money on these classes, which don't earn them any credit, and so they never get their degrees as a result. In your speech, you say institutions that use personalized learning software for remedial education see their completion rates double. How does this work? Why is this a problem that software has been better at solving than traditional methods?

There's a boundary between high school and college where the all-access colleges make you take an exam as you come in. And depending on what your math score or your reading or writing score is, if it's low enough, then you get placed in the remedial class, and they re-teach you everything. They don't tweak the results you got and say, "OK, you're missing this part or this part." It's just a binary "You're OK, go ahead" or "You have to get in the class." And so that's one of the reasons we have such high dropout rates in higher ed. If you use a personalized tool, you'd sit down and it would sort of figure out, "OK, fractions, variables, graphing." In math there's a few hundred concepts and it would sort of see which ones are weak for you.

And then you could move on to the normal courses in a variable amount, instead of taking a whole year for the remedial math. A lot of people could do it in even a month or two! And the idea of personalized learning is you always know yourself where you are on a topic, that you have the sense of what the tasks are, how much there's left to do to achieve certain levels. So there is more personal agency.

For the students who are falling behind, they'll get more time with the teacher. And it's very interesting to track, in a classroom like, that some kids go faster than others. But you have some variability where there'll be one topic where some kids are going fast, and then different kids will have a hard time on another topic. It's pretty obvious that in the one teacher, 30 kids-type classroom, there's a lot of boredom in there.

So far, there aren't many standards around what "personalized learning" software should look like or how it should work. Do you expect that a standard model will emerge over time, or will it always be in flux?

Well, I think there, like you see in the textbook market, you end up with a few products with fairly good share. It used to be what you learned every year was different in the 50 states. And now we have this thing called the Common Core, which is not a curriculum, but a common set of knowledge that a 6th grader should have, that a 9th grader should have, and so on. And that means that somebody like a Khan Academy, who is trying to be a website that's a resource that helps out with all of this, when they talk about 6th grade math, it makes sense to kids in all the Common Core states exactly what's there, the way they do the notation, the way they order things.

Today I sat with a lot of the venture capitalists who work in this field. The foundation supported a number of those. Our role in this is to facilitate the market, to make sure that if the product is good, there is really proof that it's good. We're just trying to drive toward the outcomes. There will be a mix of free software and paid-for software in this. The exact balance of that in different areas — it's not clear what that will look like.

Your speech touched on how noisy the field of "ed tech" is right now — tons of companies are building things, and if you're a teacher or administrator, it can feel totally overwhelming. How do you make the case to teachers and administrators that personalized learning is something they should pursue now, instead of a few years down the road when the field is more mature?

Well, the biggest single barrier was that telling kids to use to use digital devices outside of the classroom might discriminate against kids who might not have an internet or PC access. Now both the penetration rates of tablets and PCs are much higher, and some of this can be done on the phone screen. Deciding which things you need the big screen for versus which things work off the small screens — that's an area of interesting activity.

Like if you're just practicing vocabulary on something like Duolingo, that works really well on the phone. If you're watching a history lecture, maybe. Reading a textbook, maybe. I like big screens, but we gotta get to all the kids. In the US at least, assuming you have a smartphone, it works pretty well. So, the biggest barrier has largely come down.

Discussions of personalized learning generally go out of their way to emphasize the importance of teachers. And yet in the personalized learning classrooms I've visited, teachers play a much different role in the classroom than the ones I grew up with. Teachers unions can be fierce opponents. How do you get them on your side?

Well the world is not that different. You want to get rid of the drudgery pieces, like creating math homework and grading math homework. You know, teachers want to get in and help individual students, and so I don't think it's really a different skill set. Eventually the schools of education will expose people to these new learning models as they get really pervasive. So you'll come in ready to go with these things. There may be a generation of teachers that doesn't adopt these things. But certainly the younger teachers coming in, they've chosen to be in education, and they know this is part of it. For the K–12 teacher, it's mostly eliminating the stuff they are not that enthused about doing.

As you move up into higher ed, it gets more dramatic. Should big lectures [still] be done, with each school doing them differently? Or are you just licensing that, like you used to with textbooks, and so those big lectures are just kids watching video online? That's, in a sense, a little more threatening, and the question of whether each institution should do its own lectures really does come up. That would be a big change to get rid of that. But there's a lot of cost pressure [on schools]. And some of the super high-quality lectures, like Learning Company, are way above average. And so will people move to that? Will people resist that? That's going on right now.

HeroImages | Getty Images

Lately the tech world has been enthusiastic about artificial intelligence in general, and bots in particular. Microsoft, as you know, has recently made a significant investment in bots. What role do you see for AI in education?

Take writing — some people scoff at it, but there's early work to give people feedback about their writing that in the years ahead I think does have a lot of promise. It is very intensive to take a piece of writing and give somebody feedback to help them be a better writer. It's not like mathematics, where you are checking against this black-and-white thing. So AI tutors, that's one of the verticals that will be played with, is this whole dialogue richness.

Then you can have vertical tutors where, if you're confused about a concept, it's another level of interactivity. Today's interactivity is ok, I answered a few questions wrong, so then it repeats the lecture. [With an AI tutor,] I can engage in a dialogue. You know, part of the reason I'm so willing to tackle new subjects is that for each of those subjects, if I get utterly confused, I know somebody I can send an email to, and they'll straighten me out. If you can't have somebody straighten you out, and say, is this as complicated as I think? Am I missing something here? Then you're probably not as ambitious. For a lot of subjects, as they get older, people are not willing to take that learning risk where they are confused. The idea that you could talk to a [virtual] advisor that would understand different misconceptions and arbitrary linguistics around it, that'll certainly come in the next decade. And they'll be a very nice supplement. We already have online a bit of an ecosystem where you can go get a math tutor for a certain number of dollars per hour. But the beauty of this is it could be completely free.

By The Verge's Casey Newton. Read the original article here.