But in 2020-2021, not only the technology itself has become more mature, those who want to apply AI in practice have “matured up” as well. Teams have come to government departments that understand how to apply AI technologies and get a meaningful result, and not just a worked out hypothesis. As in the case of the refinery, we are talking about complex and at the same time quite special processes.
For example, you can automate the process of registering the transfer of ownership of real estate. Here, only on a set of two documents – an agreement and a power of attorney – you need to carry out more than 60 logical checks: check the participants in the transaction, the powers of authorized persons, whether the terms of the contract coincide with the period of validity of the power of attorney, and much more. This is a fairly serious routine workload that takes a lot of time.
Read on RBC Pro
Recently, Rosreestr also presented the results of the Smart Cadastre pilot project. The service automatically detects the absence of information about objects in the Unified State Register of Real Estate (EGRN). 2.5 times more unaccounted for objects were identified compared to previous periods. The increase in budget revenues from land tax alone amounted to 211 million rubles. in year. At the same time, it was calculated that the system found unaccounted objects about 2 thousand times faster than in the manual mode of entering information.
Since all types of state activities – administrative, licensing, registration, control and supervision – are, first of all, work with incoming flows of documents, the task is to automate this particular process – transferring data to the information system of departments. Modern AI has learned to “understand” the document in much the same way as a person does – by comparing various blocks of information with each other. Such a digital assistant can distinguish where the contract refers to the buyer and where the seller is, and check other parameters. At the current stage, systems of this class are able to take on 60-80 and even 95% of routine work and checks, leaving the final decision to the person.
To implement a project to develop a departmental AI-based digital assistant, you need to go through five steps: choose one of the document-intensive processes, conduct business design of this process, structure information, collect a primary data array and train a neural network complex on it, and, finally, after all To do this, systematically increase the functionality of the AI system by marking up and retraining. Then you can scale the system to other document-intensive processes, connect it to the processes of interdepartmental interaction.
One of the examples of already operating services based on AI elements in government activities is the call centers of the unified service 122. These are voice virtual online consultants, chat bots or an interactive voice menu, which can give a full automatic answer to some of the requests of citizens, and some – quickly transfer to call center operators. Elements of artificial intelligence have been introduced in 60 regions. Services allow you to increase the speed of service provision and the number of calls that an operator can receive. Deputy Prime Minister of Russia Dmitry Chernyshenko notes that in the near future it is planned to introduce unified technological solutions for organizing the work of 122 service call centers: the Ministry of Digital Development and Rostelecom have already prepared relevant proposals. Also, according to him, on line 122, the possibility of organizing a callback from the operator is being worked out. In case of a missed call, the system itself will find such calls and add them to the operator database for subsequent calls. This is an example of effective interaction between a machine and a person.
The main thing is to teach neural networks to selectively use the available information, to treat it critically. We constantly see examples of how AI should not be trained. One of the recent ones is an attempt to train neural networks in the diagnosis of COVID-19 using MRI images of the lungs as an example. At the entrance to the car, they simply showed where the pictures of healthy people were, and where they were sick. These parameters were clearly not enough for training. As a result, machine intelligence took a certain text font as a risk factor, which was used to sign pictures of patients with a severe course of the disease, and began to find other “wrong” patterns.
The issue of teaching AI is a matter of teacher competence and well-labeled data arrays. The main postulate is that data should be marked up as fully as possible, with the participation and control of a person. If the neural network does this automatically, without a critical eye, the learning result will leave much to be desired.
That is, we are still talking about the concept of a digital assistant, and not an independent AI like the fearsome Skynet. Unlike humans, machines cannot, on the horizon of 15-20 years, realize themselves in the multidimensional space of uncertainty that humans face on a daily basis.
This is especially important to take into account against the background of the fact that in Russia 2020-2021 can be called an intensive period of digital transformation in the state.
By 2024-25, the government plans to obtain significant results from the current stage of digital transformation in all key sectors of the economy. At the same time, a special bet is made on AI. This is stated in the Digital Economy national project, the latest strategies were approved at the very end of 2021. The number of state-funded places for IT areas in universities is increasing, CDTO educational programs for civil servants have been formed. In line with the increasing importance of technology, IT costs are rising. In 2020, the total revenue of the TOP-50 IT vendors in the public sector increased by 30%. The government is launching subsidies and preferential lending programs for breakthrough AI projects.
“According to experts, AI technologies will increase the efficiency of digital transformation by 6-7 times, and the speed of obtaining public services thanks to them will increase 10 times by 2024,” Deputy Prime Minister Dmitry Chernyshenko noted in November 2021 at the AI Journey international conference 2021.
The state is turning into a digital platform so that citizens and organizations can comfortably and quickly receive public services. This is a specific area where, with the help of AI, a measurable result can be achieved in the near future. Digital assistants are definitely the IT trend for 2022, and maybe the next few years.