The Coolest Kickstarter I'm not Backing

If you’ve read Neal Stephenson’s The Diamond Age, you’ll no doubt remember the book within the book. "The Young Lady's Illustrated Primer" was one of the most central characters that isn’t really a character. If you haven’t read it, I’ll do my best to give a spoiler free idea of what I’m talking about. Set in a nearish future, young Nell is the daughter of a druggy mother and a absent father, living in a slum. Her future seems bleak, with little prospects of even getting a decent education, when her big brother brings home a book he stole off one of the other characters. The book turns out to be a prototype of an interactive device capable of discerning her abilities, likes and dislikes, and keeping her interested while it teaches her to read and then much much more.

It was a lovely vision. As a bookish person, I was enthralled by the idea of a book that could tell me any story and teach me any subject. Doing so at my own pace, and even being able to anticipate my educational needs. It was a lovely story.

Say hello to you cute dino friend who's fun to be with.

Say hello to you cute dino friend who's fun to be with.

Fast forward to today, when the folks at elementalpath have created the kickstarter project Cognitoys. It’s a small toy dinosaur that can talk to kids. This is not just a set of canned phrases like a Tickle Me Elmo. This dino is tied into IBM’s Watson API. In the video we see it learning the child’s name and favorite color, quizzing the children, and telling a joke. The heartwarming moment came when a child asks the dinosaur to “tell me a story.” The figure replies with “How about we make a story together.” It’s hard not to get excited and maybe a little choked up by that, but it’s also not clear at this point how much is scripted by the creators and how much is the AI really interacting with the kids.

The unit can also provide the Siri and GoogleNow-like ability to answer questions like “How far is the moon.” I’ll admit that I was disappointed that all the answers used imperial units like miles. This just showed that these were answers shaped by the american creators rather than an AI giving the best answer to prepare a child for life in a global economy (where 96% of the people use metric).

If the dino fulfills its potential it’s going to be a game changer. With Watson behind it, there’s no reason to think that it couldn’t do things like

  • Learn each child’s level and preferred learning pace, adapting as the child develops
  • Learn each child’s personal preferences and best way to keep the child motivated to learn
  • Develop population models, finding unexpected successes and failures in various teaching techniques that would be expensive to test in the classroom.
  • Diagnose any special needs based on direct data from hours of tuition rather than the relatively short classroom observations that human psychologists must rely on.

That last one could be very disruptive. There is really no reason to pre-bias Watson with the diagnostic pigeon holes that we have created children’s learning difficulties. If these are real, it should discover them for itself. My suspicion is that we’d actually see are far more numerous and much more nuanced classifications that only map very approximately onto our current ideas of special needs.

Alas. I don’t think I can justify the expense. Partly, it’s because of my suspicion that this is going to turn out to be the Apple Newton of smart toys; the great idea that was just a bit too far ahead of the technology. I worry that what we’ll really get is an algorithm with a set of preconceived educational theories crammed into it that won’t be sufficiently flexible to deal with the unique set of strengths a weaknesses each child presents. One clue that this might be the case is the promise to give “age-appropriate” answers. As the father of a second grader who is a bit behind his classmates in reading, but familiar with some high-school level math concepts, I’m not convinced that an age-appropriate curriculum isn’t a bit oversimplified.

The real reason I’m holding off is because of what’s clearly missing from this: A visual interaction. Nell’s book could talk to her and show her things. IT could encourage her to trace out letters, teaching her to both read and write. It could show her pictures of the historic individuals she was learning about, or diagrams of the concepts. I doubt Watson’s API’s have provision for that kind of interaction, but it wouldn’t surprise me if it did very soon. I think I’ll hold out for that.