1. What does AI do best today?
Now it is difficult to find an area in which artificial intelligence is not used. Moreover, both where people used to do it manually, and in fundamentally new areas. After all, it is better when AI brings something new, previously inaccessible, and does not replace a person by automating routine operations. In such scenarios, large amounts of data are processed, response times are significantly reduced, or other factors arise that a person cannot cope with on their own.
“ One can note the following tendency: if earlier banks mainly experimented with AI, now the technology is actively penetrating the real sector of the economy.”
Modern cases most often relate to production tasks. Moreover, if a few years ago these projects were rather the subject of a scientific article, now they play an important role in corporate processes. A good example is predicting the failure of typical equipment. In this case, it is not enough just to predict the failure of a unit or part. It is necessary to classify such failures and ways to repair them, learn to predict those of them that should be preventively eliminated. Moreover, it is important to be able to predict failures with a margin of time in order to have time to react.
Another trend is the growing interest in optimization problems. We see that it is the optimization cases that attract the special attention of our customers.
This is the first artificial intelligence system that can discuss complex topics in a lively discussion. The goal is to help people build convincing arguments and make informed decisions. Project Debator analyzes massive texts, builds a well-structured speech on a given topic, expresses it clearly, refutes the arguments of the opponent. Project Debater helps people reason by providing compelling, fact-based arguments and limiting the impact of emotion, bias, or ambiguity.
2. What is the situation in the development of technologies using AI in the world now? What areas are the focus of special attention on the company and developers?
Speaking about technologies using AI, it should be understood that the success of AI implementation depends on many factors, including the necessary server hardware, software tools, and the competence of employees. And at the forefront is the availability of well-collected, classified and consistent data. In all areas, we see tremendous development from both users and vendors.
“If we talk about the focuses of development, then, first of all, this is ensuring the explainability and ethics of AI, as well as reducing cognitive distortions.”
3 . How are things going in Russia? Where are we “ahead of the rest of the world”, and where are we lagging behind? What do domestic companies need to be competitive in this area?
Based on communication with customers, business partners and the academic community, Russia does not lag behind global trends, and in some areas is even one of the leaders in the implementation of AI. Given the structure of the economy, AI is being implemented primarily in geological exploration, financial institutions, and the oil and gas industry. But we know about quite successful pilot projects and implementations in medicine and education.
4. Everyone has heard of the use of AI in the field of medicine, the production of drones. What about other industries?
As mentioned, it’s hard to find an industry right now that isn’t adopting AI or considering doing it. For example, the financial industry was one of the first to use AI to minimize the risks of information leakage and hacking by automatically detecting anomalies in the access and use of corporate information systems.
5. If we talk about such industries as education, management and other humanitarian areas – is there a place for AI in them?
There are both general use cases for AI and industry-specific ones. In management, for example, it is often necessary to make a decision based on a large amount of input information.
“Nowadays, AI still often plays the role of an advisor, and makes decisions in those areas that are well regulated and require an automatic response, for example, infrastructure monitoring or cybersecurity.”
As for the humanitarian areas, over the past year, a significant part of training programs has moved to online platforms, and many educational institutions are faced with the problem of a lack of AI solutions for drawing up an individual program, automating task checking, speech synthesis and voice-to-text translation.
6. How to prepare the infrastructure in the company for the implementation of products based on AI?
When we talk about preparing the infrastructure for AI, we mean the main stages of AI implementation:
- task definition and goal setting;
- collection, organization and storage of data;
- training machine models;
- piloting the project and analyzing the results;
- scaling and commissioning.
Each stage requires the appropriate IT infrastructure. For collecting and storing data – productive and capacious data storage systems (DSS), for organizing, processing and training models – specific equipment using a GPU, TPU or CPU. Industrial deployment requires high-performance and resilient servers with fast interfaces, such as IBM POWER Systems.
The Mayflower is a fully autonomous, AI- powered marine exploration vessel. Its purpose is to collect critical ocean data. Scientific equipment is installed on board. In the role of the ship’s captain – AI. By assimilating data from a variety of sources, AI continually evaluates its route, status and mission and decides what to do next. Cameras and computer vision systems scan the horizon for hazards, and streams of meteorological data reveal potentially dangerous storms.
7. What kind of restructuring do the business processes of a company planning to implement AI need?
Changing business processes is always a creative process, as well as a strategic choice between digitalization and digital transformation. Whether there will be a need to change business processes or not depends on both the AI implementation scenario and the company’s strategy. It is often more effective to integrate AI into the current business process so that other participants do not feel inconvenient. But sometimes it is necessary to completely change the approach and even think about changing the level of industry standards or the regulatory framework.
When it comes to skills in AI ethics and data science, an acute shortage of skilled workforce is certainly hampering development at the global level. There is no quick fix, but companies can invest in employee skills and, again, use AI to personalize learning, as IBM does.
8. What are the weaknesses of AI and products based on it? What can and cannot he ever be?
If we omit the talk about the lack of strong or general artificial intelligence today, then the weaknesses of AI are now largely determined by the level of ethics or the ability to build objectively fair models, as well as the quality of the data used.
9. Artificial intelligence is a very powerful tool. What are the key approaches to the ethics of using AI now?
IBM identifies the following key approaches to AI ethics:
1. The goal of AI is to complement human intelligence. IBM is of the opinion that AI will help make each of us work more efficiently, and that the advances of the AI era should be available to many people.
2. Data and analytical materials belong to the creator. IBM customer data and analytics are owned by customers.
3. New technologies, including AI systems, must be transparent and reasonable. IT companies need to have a clear understanding of who is training AI systems, what data is at the heart of the training, and most importantly, what the algorithm recommendations are based on.
“RoboRXN Remote Chemistry Lab is built on RXN for Chemistry technology and is managed via the cloud. It is based on an artificial intelligence model, which constantly studies the characteristics of chemicals and recommends the correct sequence of operations to “prepare” a specific target molecule. The technology helps chemists predict chemical reactions with 90% accuracy.”
10. What is your forecast for the development of AI in the next 20 years?
On the one hand, 20 years is a long time. But in reality this is not at all the case, because AI systems have been studied and developed since the 50s of the last century.
“Given the current dynamics, the line between the real world and its digital counterpart will continue to blur.”
Technologies will mix: quantum computing, cloud computing and AI solutions will merge into single hybrid platforms that can be used to create, test and deploy custom scenarios. Now for the successful creation and implementation of AI, a qualified team is needed, but in 20 years this will be taught at school. Most likely, the development of AI will go in the direction of deepening and personalization. Everyone will have their own personal AI assistant that matches the nature of work and private life. Thus, artificial intelligence will become less artificial and more intelligent.
