111–120 of 325 posts
Not sure what to think about it. Seems awesome at first glance but then the examples are skipping songs and hand waving pokemons. Feels a lot like a solution looking for a problem.
An increased level of spatial awareness for phones will be huge in the coming decade. However, it will almost certainly be a result of sensor fusion between a Soli-like radar sensor, a FaceID-like ToF sensor, enhanced positioning and pose detection, RGB cameras, microphones, and a lot of ML to assemble a comprehensive picture of environmental context and user intent.
Radar is one more piece of the puzzle in building products that can read the same cues we naturally use to communicate with other humans: Imagine, instead of telling a voice assistant “Alexa, turn down the volume,” where you have to use a phonetic trigger, and all the system has to go on is audio, something more natural: You look in the direction of the hardware, say “turn it down a bit,” and make a pinching gesture with your hand. The system can assemble all these pieces (you were looking at it, you spoke in its direction, you gestured) and, with a sufficiently-trained neural network, make a more conclusive determination of your intent.
Earlier quoted context omitted.
How is Soli any different? I know that if I put my hand in this spot and make the motion of turning a dial then it will turn down the volume, and I can hear the volume go down as I do it, whats the difference? Or if I tap the air in a specific way, get an audible tap noise, and know I've completed a preset action (index finger tap = set timer for 5 minutes) its the exact same thing. How is sound feedback not as good?…
Hypothetical scenario: User pays the initial cost of engagement - Atleast 2 seconds to unlock the phone. After that, all actions user performs are in the "Engaged State". Imagine if you have to press a button every 3 mins but you have to pay the 2 second cost of unlocking the phone, searching for the button(which can be memorized) and then getting visual/sound feedback. Compare this with a Remote Control that is sitt…
In my mind the only difference between a button on a remote and a "button" that sits in the air above my coffee table is reliability, which will be solved in the future. I know pressing button x does y, it makes equal sense that moving my hand like x does y. Feedback could be anything from a puff of air, tiny light flash, small 'ding' sound, etc.
Earlier quoted context omitted.
It will just join the graveyard of gimmicky vr/ar motion controllers. Leap motion, Kinect, etc. Those can track precise skeletal gestures too. I think the differentiation here is low power always on and attached to the phone with high field of view.
The Kinect caused a huge wave of innovation. It was a convenient and low-cost source of RGBD data, and many robotics labs got a few when it was released.
I have personally built stuff with KinectFusion before so I do know how powerful it is.
Earlier quoted context omitted.
It will just join the graveyard of gimmicky vr/ar motion controllers. Leap motion, Kinect, etc. Those can track precise skeletal gestures too. I think the differentiation here is low power always on and attached to the phone with high field of view.
The Kinect caused a huge wave of innovation. It was a convenient and low-cost source of RGBD data, and many robotics labs got a few when it was released.
Earlier quoted context omitted.
How is Soli any different? I know that if I put my hand in this spot and make the motion of turning a dial then it will turn down the volume, and I can hear the volume go down as I do it, whats the difference? Or if I tap the air in a specific way, get an audible tap noise, and know I've completed a preset action (index finger tap = set timer for 5 minutes) its the exact same thing. How is sound feedback not as good?…
Sound is OK for a feedback of success. What we also need is visible affordances -- signals of what can be done, when, where, and how. Also, what I hate the most with these gestures is fearing that I'll get it close but not quite, and something else (or nothing) will happen. With a button, you can feel it and immediately make micro-adjustments while pressing to ensure success. This can possibly be done with sound, but…
Part of that is the two games on the announcement page which show you examples of gestures. Doing gesture x accomplished y in the game, so maybe in the next app you open doing gesture x will accomplish something.
Earlier quoted context omitted.
The Kinect caused a huge wave of innovation. It was a convenient and low-cost source of RGBD data, and many robotics labs got a few when it was released.
Could you expand on the result of said innovation? Where would I see Kinect driven innovation in everyday life?
Those growing up with Remote Controls of the kind: https://i.imgur.com/AIqz63k.jpg After a few days, the user develops a muscle memory of sorts. User doesn't even have to look at the controller and all actions (and feedback) are executed through the tactile interface. From cockpits to nuclear power plants to home tv remote control, there is absolutely nothing that replaces physical buttons, encoders, sliders and togg…
I know it’s one data point but I never developed muscle memory for remotes the same way as, let’s say, Emacs. I always stumble and press the wrong keys. More importantly the interface on the TV is convoluted. I don’t want next channel - I usually want to go to a specific channel. With, say, Apple TV it means scrolling through a bunch of apps - much faster. Or YouTube tv via chrome cast - scrolling through the guide o…
I like remotes like the Apple TV remote and the Roku remote (except for the fact that the Roku remote is a piece of utter garbage and I want to have strong words with the designer) because they focus on the essentials and let you do everything else through menus.
(The Roku remote is an utter piece of trash for a couple reasons. First, it chews through AA batteries like it’s going out of style. Second, it is not responsive—there is a agonizing delay between when you press a button for the first time and when the action takes effect. My guess is that these defects are because it uses WiFi to communicate with the device, and I can’t understand why anybody thought that was a reasonable technology to choose.)
The first is learnability. A big problem with gestures is that there is no clear affordance as to what kinds of gestures you can do, or any clear feedback. For feedback, one could couple Soli's input with a visual display, but at that point, it's not clear if there is a big advantage over a touchscreen, unless the display is really small.
The second is what's known as the Midas touch problem. How can the system differentiate if you are intentionally gesturing as input vs incidentally gesturing? The example I used was the new Mercedes cars that have gesture recognition. While I was doing a test drive, the salesperson started waving his hands as part of his normal speech, and that accidentally raised the volume. Odds are very high Soli will have the same problem. One possibility is to activate Soli via a button, but that would defeat a lot of the purpose of gestures. Another is to use speech to activate, which might work out. Yet another possibility is that you have to do a special gesture "hotword", sort of like how Alexa is activated by saying it's name.
At any rate, these problems are not insurmountable, but it definitely adds to the learning curve, reliability, and overall utility of these gesture based interfaces.
The "dial" gesture is incredibly subtle. Kudos to them if it works reliably. MKBHD doesn't seem convinced in the efficacy of the sensor* from his Pixel 4 video: https://youtu.be/sKJ4i7p-o-4?t=326 * from Analemma_'s comment below, the Pixel 4 doesn't seem to be running the full blown chip