OptiKey – Full computer control and speech with your eyes

Mmmm... that was a great deal slower than me using the latest Android swiping support, even counting errors. Can't speak to the other mobile tech stacks, but if you're typing on a phone is slower than that you might want to poke around, you may be missing a technology already installed on it.

(Not a criticism of the excellent work here. It's a fundamental bandwidth problem.)

Concretely: In the time the demo took to write "Meet OptiKey. Full computer control and speech using only your eyes.", I wrote in an Android text area: "I am just typing words as quickly as I can in the android interface. I'm even having to think a bit about it more that I've written so much. uh, hello world? and other stuff. I'm still going and going and going." Including the missing "now" between "more" and "that", and you can't see it but I had to correct 3 words, too. That's about three times faster on the phone I have in my pocket. And let me just say... that is impressive performance for a free eye tracking suite, in my opinion, to still be that fast with just eyes.

this is so mind blowingly good.

If we're talking about what's optimized for eyes and the movement of them, you wouldn't even want a keyboard layout, you'd want something more like a dart board or oval shaped dartboard where the spaces are letters and the most common letters are towards the center. QWERTY is fine for someone familiar with it, like an experienced typist whose already mapped those keys mentally, but I expect someone with a lifelong disability would have no true familiarity with QWERTY (since they can't type) and would have to learn any layout from scratch anyway, so you might as well start from scratch.

Here is an idea: Tell me if this is stupid:

Assume you have a bunch of HUDs/AACs/Glasses/Whatever that are using this tracking tech. Assume that they are ONLY looking at the real world, not some online data/webpage etc...

There is a camera that either is fwd looking or also 360 looking.

Use the tech to eye track exactly everything that MANY people are looking at to train an AI as to what items in the real-world are important.

i.e. "the ground exists and we know its there, thus its priority in information is low" however "these signs that are being looked at have a higher context priority, and require understanding"

By doing this on a fair scale... you could train AI to use the information of "whats visually important to human navigation" to train them to navigate. This augments what other ML/AI stuff has already been going on...

I do not know if this is basically how self driving cars were developed -- but now that you have a seed of this tracking tech in open source -- it could blossom.

This is a good point - how do copyrights protect intellectual property from open source alternatives? I think that this is a really good thing to open-source, but the law can be tricky to navigate. Almost like you need a degree in it...