The artificial intelligence (AI) creation is a water-cooler topic today. The president of Russia V.Putin, the president of Sberbank G.Gref, the president of Rostelekom M.Oseievsky are talking non-stop about the necessity to start working at it. But it seems they are 5 years late.
Still, very little is being done in this direction in Russia, especially, on the part of the government. The absence of speakers of government companies at top conferences on this topic is indicative of it. Until now, the Endurance startup has been the only representative of Russia addressing such conferences.
In the USA they speak less but do more, and not government institutions but private companies, such as: Google, IBM, Microsoft. They share their achievements whenever possible. Perhaps their solutions are far from perfection, but the developers do their most to advance their technologies and make them available to all. This is what matters!
At the latest conference Bankir.Ru at IIDF (Internet Initiatives Development Foundation), 22 Aug. 2017, there was no representative of the government sector, and the organizers of the conference didn’t address it on the topic. Why?
All talks without deeds do not impress the world very much.
I like it very much that IBM, Google, Microsoft demonstrate their solutions and show real business cases. This has credibility.
Thus, what have these three companies: IBM, Google, Microsoft done in the sphere of AI?
They have been working at the services, which will understand normal human speech. This is what Google NLU, IBM Watson, Microsoft LUIS are namely intended for. You might ask what for? It is the common truth that humans are humans thanks to speech. So, the speech recognition is one of the most important elements in the process of the AI creation.
We can assert with confidence that by having taught the computer to understand the human speech, we will open the first bolt on the door called artificial intelligence.
Attempts to teach the machine to recognize speech have been made for a long time. But this became possible only now, owing to the powerful enough computer clusters and a sufficient amount of data. A number of large US companies are developing services to recognize the intentions of the speaker. In other words, they try to teach the machine to recognize not just separate words, but the meaning of what has been said, the context. This problem is already being solved very well for English. There are entire groups of dialogues on the basis of which the machine can communicate with a person in a fairly understandable human language.
To date, it seems to me that Google has come closest to solving this problem. Google is the first to introduce a good and high-quality service for speech recognition (this is what we use when we say «Ok Google»). They have enough data to analyze. More than 6 billion requests a day pass through Google, which, you must agree, is a very impressive figure. Analyzing these data, developers learn to understand what the user really wants or is looking for, and train the machines to understand these things too. It is very important.
Such services are the first vital step on the way to creation of artificial intelligence.
Until recently, Russia has not had a single similar service for the Russian language.
At the beginning of summer 2017, ParaFrazier introduced a unique FREE API, with which you can process a text: analyze the meaning, look for synonyms, select word forms, etc. This is the first service in Russia, beginning to solve the global problem of text analysis.
The project has not financing ether by a large company, by the state, or by private investors. ParaFrazier is a small group of enthusiasts, who understand that without this it is impossible to approach the world singularity.
Obviously, the project, as it develops, will attract more and more users.
Now more than 100 people a day uses the service, which indicates its relevance.
The service developers plan to learn to define complex meanings, topics and tone of the query, do intellectual paraphrasing of texts and much more.
Prior to this, Abbyy had tried to solve a similar problem, but, unfortunately, their Compreno service did not start working in the public domain. When I tested it, the results were somewhat discouraging. I heard similar feedback from many large companies that used this service.
The company MDG declares the solution of the problem in a narrow form, but again, it does not provide an open API for developers.
Therefore, we cannot but hope that the ParaFrazier project will work in full force in the near future, and Russia will step closer to the «digital economy» and create artificial intelligence.
The Endurance project supports the ParaFrazier service with their solutions and plans to launch several large joint pilots for intelligent text recognition and processing.
With their help, we expect to process complex knowledge bases of large companies for uploading them to ChatBots.
