How AI can be used to treat dyslexia

Dr. Coral Hoh, CEO and founder of Dysolve, explains how AI can be used to treat dyslexia.

Transcript

00:00Imagine that you need an expert for something like this, for diagnosing and correcting a condition, a brain condition.

00:08So what we've done is really put the expert in the machine.

00:12And so then what does the expert need to do?

00:15Let's just say it's a child that has to be diagnosed and the intervention has to be given to the child.

00:21The expert has to interact with the child in some way and observe what is going on.

00:27So then Dissolve AI creates a game, generates a game or builds a game and gives it to the child through the device and the child plays the game.

00:37So now there's a way for Dissolve AI to observe what happened.

00:42Was the child able to do this language activity or were there errors?

00:46Now if it's correct, then Dissolve AI will give another game to probes perhaps some other area.

00:53But if there's an error with this particular activity, then Dissolve AI wants to figure out what is going on and probe some more.

01:02And if the result is kind of vague and fuzzy, Dissolve AI has to think, hmm, this is interesting.

01:11What am I going to do next?

01:13But unlike me, where I have to sit there and think about it for a while, Dissolve AI has to act immediately.

01:20So Dissolve AI will just generate these game after game after game so that the child doesn't have to wait.

01:26And for language processing, it's got to be done super fast.

01:30So the question is, why do we build a computer expert system?

01:34Why wasn't human experts sufficient to do this, right?

01:38So there are three factors.

01:40One is complexity.

01:41The other one is speed.

01:43And then the third one is capacity.

01:45So let me handle one at a time.

01:47Complexity.

01:48This is a language processing issue.

01:51And so we have to consider the linguistic system of the brain.

01:55And that's like a super complex system, like a large computer operating system with millions and millions of lines of code.

02:03So any part of the system can have these inefficiencies to different degrees.

02:08How is a human going to find them?

02:10That's one.

02:11The other one is, as I said, you can't make the child wait.

02:15So it's a matter of speed.

02:17And natural language is processed super fast in the brain, under a second, hundreds of milliseconds, in parallel.

02:24Many, many things happening at the same time.

02:26So there's no way for a human to record it and to calibrate it or to calculate it.

02:33The third one, of course, is scale.

02:36Even if one person can handle one child, how about the 30 million students around the country?

02:42At least 30 million with this kind of condition.