When Apple first introduced Siri, the world of technologies started up - and many said: “Well, finally, we will live in the sci-fi future, where our little pocket robots will do all the work for us, and we will shout at them”. In practice, Siri turned out to be a bag with everyone. Intel is currently working on a proprietary version of Siri, but avoids the use of the cloud, because it will make the product much better.
The voice recognition system has come a long way from the days when something unintelligible was used, such as ViaVoice from IBM; now (if not us, then at least over the hill), you can say a few words to your computer in your pocket and he will buy a couple of games. If you have an Xbox One, you can sit on the couch in the dust of chips eaten and order your console to turn on night basketball or download the three seasons of Strelka from Netflix without even touching the joystick. When the voice recognition system works, it is incredibly practical. However, when it does not work (and this happens often), you start screaming at an inanimate object, wasting time, instead of just pressing a couple of buttons that never failed. Siri, in principle, works well. But Intel believes that Siri could work better, and this is not related to the expansion of lexical stock and polyglot.
With Jarvis, a new headset that is being developed and named after the Iron Man Butler, Intel plans to remove the cloud from the equation, reducing the time required for the speech recognition system so that it understands your vague commands. The main voice recognition platforms work in such a way that they compress your voice command and send it to a central server. Computers on these servers translate a voice command into text or command, and then send it back to the device. Obviously, this process can be incredibly slow due to external factors that can not always be controlled, for example, the speed of connection to the Internet. By doing everything to leave client-side processing, Intel can reduce the time it takes to send a voice command to the server.
By partnering with an unnamed third party, Intel has created a wearable device that processes voice commands without any servers. Currently, Jarvis is a headset that sits inside your ear and connects to the phone wirelessly. Because of this local interaction, Jarvis does incredible things and works even when your Internet has ordered to live long - what else do fans of the subway dream about?
A more technically savvy crowd will tell you that even when voice recognition works flawlessly, working with it is not as convenient as pressing buttons. This is because most voice recognizers do not respond as quickly as you can press a button, even if the sound spreads faster than the movement of your fingers. Thus, the goal of the ideal voice recognizer is to respond immediately after the command, and this is exactly what Intel is going to do.
Currently, Intel is working on selling technology to phone manufacturers, where it will be able to express themselves in the best possible way. Of course, as soon as the issue of fast data processing is settled, developers will still have to develop software that will never get us wrong and will end the sentences for us.
The article is based on materials .
Comments
Post a Comment