⬆️ ⬇️

Home robot: from idea to product

Since ancient times, people wanted to talk with the outside world. Talking things surrounded him in fairy tales and legends. The evil queen communicates with the talking mirror, Gollum complains about the hardship and adversity of being the One Ring, Tamerlane spoke to the stones and gave them names.



All this is because the voice is the most natural way of communication for a person, verbal communications are embedded in our genes. And today we are very close to embodying the ancient dream of mankind with the help of high technology.



In Cubic Robotics, we believe that the fastest way to the “talking world” lies through the creation of a universal voice operating system, which on the one hand could be installed on any device. On the other hand, such a system could be developed by independent developers around the world. Such is the VOIS (Voice Operation Intellectual System) voice operating system developed and patented by us.

')

image



VOIS is a universal voice operating system, with which you can make almost any subject speaking. To talk means not just to take voice orders, but to really understand a person, to support and even initiate a conversation. This is the same artificial intelligence that is inside our development - the robot assistant Cubic (we described it in detail in the last article).



Also, on the basis of VOIS, talking cars, intelligent assistants for smartphones and wearable gadgets are already being created and designed.



Work principles



You need to be honest with yourself: you can hardly create real artificial intelligence in the near future: you need huge money, the computing power of huge laboratories, and the best minds of the planet. However, the existing technologies allow to create an operating system with AI elements.



From a technological point of view, VOIS is a programming language, the core of the system, and a set of applications that control various speech functions.



The basis of the principle of VOIS is the ability to understand the context of the conversation, as well as the emulation of human conversation. You ask a question and get an answer, not a set of links found in a search engine.

In fact, from the inside, the process looks like this: as soon as the user’s phrase arrives at the server, he begins to look for options that might be suitable as an answer. To do this, we use several different engines from the classic chat bot to the statistical algorithm for searching through the base of dialogs. For example, a user asks - "What's in the street"?



There are several answers to this question. Among other things, the server will consider the following options:

- On the street 25 degrees and the sun shines.

- There are very serious traffic jams on the street, you should not drive a car.



And a search through a database of successful dialogues can produce something like: “You yourself can look, unlike me, you have legs.”



Regarding the correct answers can be quite a lot (hundreds). It is important to choose the one that the user needs right now. The algorithm for checking the context (statistical) is responsible for this.

It can show that if the database has information that the user has a car, then he is most likely interested in traffic jams. And if the user just talked about the weather (for example, asked the weather for tomorrow), then it is more logical to answer him about the weather on the street.







Defining the context is a difficult task. Partly it is based on an analysis of the existing database of correct dialogs. Some rules have to be written manually, others follow from formal logic. For example, the simplest and most effective rule: if two answers are equally good, then the one whose topic is closest to the last conversation will be chosen.



Of course, at the present stage of development, VOIS periodically gives very strange answers: many scenarios have not yet been written and are not entered into its database, so curiosities happen.







What will happen next

Today, a number of companies have already shown interest in VOIS and devices based on it: not so long ago we communicated with representatives of Samsung and KAMAZ.

Samsung met with us twice, representatives from the company's head office, who were very interested in the Cubic project, came to one of the meetings. In turn, KAMAZ proposed to create a voice interface for the onboard computer of a new model of KAMAZ.



In both cases, we have not yet reached any concrete agreements, but negotiations continue.

However, our main goals now are to launch the serial production of the Cubic robot (debugging work is now under way at a factory in China) and to create an API for VOIS, in order to make the system accessible to each developer.



image



We hope that by the end of 2014, an infrastructure will be created that will enable everyone to develop their own applications for VOIS or create talking devices based on it.



Moreover, the development of applications for VOIS is much simpler than creating such things for smartphones: graphical user interfaces are not needed.

We will describe examples of integration in our system in various devices in the next article, but for now we are ready to answer your questions.

Source: https://habr.com/ru/post/222655/



All Articles