1.) Why no mention of the Google Glass (and the «Glasshole» phenomenon) ? It would be interesting to compare their sales - I've only been able to find old and very uncertain numbers after a quick search. More about Hololens too.
2.) Large Language Models seem to be (mostly) a bad fit, considering how they are optimized at working with written text, and you're not going to get the kind of interface that allows you to write text on glasses. (Probably... though maybe combined with bracers with keyboards ?? Those are already popular in some work contexts (supermarkets).)
(Also, the legality of neural networks in general is still under question. Ditto for USian infocoms, especially GAFAMs.)
3.) If the smartglasses are made by a USian infocoms, I would expect seeing more ads, not less (well, except maybe for the high end models, which are fated to stay a small part of the market - see also : iPhones).
4.) > The shape of the human head is going to stay consistent for the foreseeable future, and whilst there is some leeway to be found in the flexibility of social dynamics, we can expect the form factor (and therefore volume) of the glasses to be approximately similar to glasses today.
I really would be careful with a prediction like that. Consider how social acceptability of people always having a cellphone/smartphone on them radically changed in only 2-3 decades. Or using headphones in merely half a century. If the experience is worth it, I would expect Hololens-shaped devices to catch on quickly (~570g compared to your suggestion of 75g as the top end !) But maybe slimmer, compare first generation mallet cellphones with the Nokia 3XXX line. (Again, see also how keypad bracers and Hololens are already a success in work contexts.)
5.) > J/h
That's a weird unit if I ever saw one, why aren't you converting to Watts ?
6.) > SOTA
(I know that it means «state of the art», but maybe avoid unnecessary abbreviations in your writing ?)
7.) > to cover a ~100° field of view (FoV) with human eye acuity of 1 arcminute, ideally 6K6K resolution is required for each eye [9]. This equates to roughly 1µm pixel pitch
What do you mean by «high resolution» ?
1 pixel / arcminute = 60 pixels / degree (aka 20/20 vision) is very commonly cited as the maximal acuity for human eyes. It's also a number that is wrong :
> NHK claimed the tests showed 310 pixels/degree are needed for an image to reach the limit for human resolution [in the context of watching TV].
(This seems to include the doubling for Nyquist–Shannon, as one might expect.)
This is roughly «11k».
(But then, «in context of watching TV» might matter, only actual testing will tell what the maximum figure is for each kind of glasses.)
(Incidentally, 20/20 vision is pretty bad : it's what a postmodern 60 years old has, after correction.)
But then, this is the top end resolvable by an average human, «high resolution», while relative, could be said to be hit long before that ! (Context of usage matters, for instance «high resolution» is much lower on TVs compared to monitors, where you're expected to interactively work with a lot of text.)
Last but not least, you don't need the resolution to be high across the whole 100° of the field of view : the human eye fovea that is responsible of most of the daylight acuity is only ~5° in size... also is almost entirely blue-blind (the situation is complicated by the eyes constantly moving, among other things). IIRC Virtual Reality headsets have already started to exploit this to massively reduce computation and therefore power needs ?
8.) I'm not sure what do you mean by «full stack», how is it different from «vertical integration» ?
9.) > Other: Everything else a standard smartphone does
That too might be shortsighted, a very different form factor probably means that successful devices won't need to replicate «everything». For instance, smartphones these days generally don't feature infrared transmission, (direct) radio listening, or physical keyboards, features that were popular on early smartphones, and that's for a very similar form factor !
10.) That triangle with smartphone might also be shortsighted, for instance people typically don't carry any more radios, walkmans, pagers, often not even a watch (whether dumb or smart) ! So the smartphone might disappear too (for instance if the computing center becomes on the back of a Hololens-like glass and/or in a bracer). After all, a cell/smartphone needs a large enough pocket or purse, and is fairly easily dropped/lost/stolen...
11.) What would an «ultimate form» or «final generation» even mean ? By definition, that would be one before the device's disappearance, by which point its usage becomes niche (see : the recently disappeared telegraph), so why focusing on that ? (And you might never know when it might reappear when the context changes, so that finality itself is always under question.)
- Yeah including info on the Google Glass would have been valuable.
- LLMs are a bad fit, but VLMs perhaps not!
- I agree in the longer term the dynamics will be much more flexible, I was referring to near term first gen smartglasses!
- Everyone uses SOTA in my sphere :)
- The 6K6K figure was taken from the references, and high resolution there is ~indistinguishable. Maximising resolution just for the fovea was a step too far for the blog post, but yes super effective technique!
- Yes full stack ~= vertical integration. What im getting at there is you can't just slap a skin on android and call it done as many vendors are currently doing.
- Yes agreed, triangle is near term as the rest of the blog is focused on first gen :)
- Final generation is iPhone X vs iPhone 1, obviously the real final generation will be ~Neuralink!
1.) Why no mention of the Google Glass (and the «Glasshole» phenomenon) ? It would be interesting to compare their sales - I've only been able to find old and very uncertain numbers after a quick search. More about Hololens too.
2.) Large Language Models seem to be (mostly) a bad fit, considering how they are optimized at working with written text, and you're not going to get the kind of interface that allows you to write text on glasses. (Probably... though maybe combined with bracers with keyboards ?? Those are already popular in some work contexts (supermarkets).)
(Also, the legality of neural networks in general is still under question. Ditto for USian infocoms, especially GAFAMs.)
3.) If the smartglasses are made by a USian infocoms, I would expect seeing more ads, not less (well, except maybe for the high end models, which are fated to stay a small part of the market - see also : iPhones).
4.) > The shape of the human head is going to stay consistent for the foreseeable future, and whilst there is some leeway to be found in the flexibility of social dynamics, we can expect the form factor (and therefore volume) of the glasses to be approximately similar to glasses today.
I really would be careful with a prediction like that. Consider how social acceptability of people always having a cellphone/smartphone on them radically changed in only 2-3 decades. Or using headphones in merely half a century. If the experience is worth it, I would expect Hololens-shaped devices to catch on quickly (~570g compared to your suggestion of 75g as the top end !) But maybe slimmer, compare first generation mallet cellphones with the Nokia 3XXX line. (Again, see also how keypad bracers and Hololens are already a success in work contexts.)
5.) > J/h
That's a weird unit if I ever saw one, why aren't you converting to Watts ?
6.) > SOTA
(I know that it means «state of the art», but maybe avoid unnecessary abbreviations in your writing ?)
7.) > to cover a ~100° field of view (FoV) with human eye acuity of 1 arcminute, ideally 6K6K resolution is required for each eye [9]. This equates to roughly 1µm pixel pitch
What do you mean by «high resolution» ?
1 pixel / arcminute = 60 pixels / degree (aka 20/20 vision) is very commonly cited as the maximal acuity for human eyes. It's also a number that is wrong :
https://web.archive.org/web/20230323210745/http://www.homeci...
> NHK claimed the tests showed 310 pixels/degree are needed for an image to reach the limit for human resolution [in the context of watching TV].
(This seems to include the doubling for Nyquist–Shannon, as one might expect.)
This is roughly «11k».
(But then, «in context of watching TV» might matter, only actual testing will tell what the maximum figure is for each kind of glasses.)
(Incidentally, 20/20 vision is pretty bad : it's what a postmodern 60 years old has, after correction.)
But then, this is the top end resolvable by an average human, «high resolution», while relative, could be said to be hit long before that ! (Context of usage matters, for instance «high resolution» is much lower on TVs compared to monitors, where you're expected to interactively work with a lot of text.)
Last but not least, you don't need the resolution to be high across the whole 100° of the field of view : the human eye fovea that is responsible of most of the daylight acuity is only ~5° in size... also is almost entirely blue-blind (the situation is complicated by the eyes constantly moving, among other things). IIRC Virtual Reality headsets have already started to exploit this to massively reduce computation and therefore power needs ?
8.) I'm not sure what do you mean by «full stack», how is it different from «vertical integration» ?
9.) > Other: Everything else a standard smartphone does
That too might be shortsighted, a very different form factor probably means that successful devices won't need to replicate «everything». For instance, smartphones these days generally don't feature infrared transmission, (direct) radio listening, or physical keyboards, features that were popular on early smartphones, and that's for a very similar form factor !
10.) That triangle with smartphone might also be shortsighted, for instance people typically don't carry any more radios, walkmans, pagers, often not even a watch (whether dumb or smart) ! So the smartphone might disappear too (for instance if the computing center becomes on the back of a Hololens-like glass and/or in a bracer). After all, a cell/smartphone needs a large enough pocket or purse, and is fairly easily dropped/lost/stolen...
11.) What would an «ultimate form» or «final generation» even mean ? By definition, that would be one before the device's disappearance, by which point its usage becomes niche (see : the recently disappeared telegraph), so why focusing on that ? (And you might never know when it might reappear when the context changes, so that finality itself is always under question.)