17 декабря 2018 г. 10:09

Artificial Intelligence: the Reality and the Future

In his interview Stanislav Voronin, Head of the BI Department at Softline, explains the capabilities of artificial intelligence, usage scenarios and covers AI-powered solutions from Softline.

- Stanislav, please tell us about some AI use cases. In which areas can it be applied?
- There were attempts of using AI in almost all domains: medicine, finances, education, business, entertainment, law enforcement, etc. In 2010s we can see rapid development of artificial intelligence technology although the neural networks had been invented as early as in the middle of 20th century. In fact, at that time people did not have the sufficient computing capacities to implement these solutions in code and now we have them at fingertips. Now every one of us can see the AI in action, for instance, by using the voice assistants in smartphones: Siri, Google Assistant, etc. These systems can analyze data, encrypt, and synthesize speech.

There is a solid commercial reasoning behind using AI technologies. In modern business situation, they bring tangible profits. For instance, Rolls-Royce, an airliner engine manufacturer uses AI technology to optimize the engine performance, predict possible outages and plan an optimal repair and maintenance program. As a result, the service interval mileage increases, the failure probability decreases virtually to zero, which, in its turn, increases the product safety. All of the above enables a business model: an airline pays not for engines but for the hour of their operation.

Some insurance companies use chips mounted in cars that collect information about driver’s behavior including driving style, the amount of time spent driving per day, the speed, the acceleration and braking dynamics, and many other things. The AI analyzes a huge number of parameters and calculates how high the probability of a traffic incident is for a specific driver. In future, it uses this data to set an insurance tariff.

Banks use similar technologies for physical person scoring. Before approving a loan, they analyze very much information about the client: credit history, questionnaire data, personal characteristics, etc. It is also possible to compare data collected in previous credit requests with the currently provided information. As a result, the system reveals the real state of affairs of such a borrower. The number of such applications is going to increase in the years to come, they will become ubiquitous and transform all fields of our life.

- How do AI systems work?
- They are based on neural networks and other methods of machine and deep learning involving the process of training models to solve specific problems. First, you need to load the input data and train the neural network to distinguish certain objects from others, compare them and predict the outcome. For example, in order to learn how to recognize the images of cats, the system needs to analyze many photos, categorizing them by characteristics and parameters. Scenarios include predicting stock prices, recognizing handwriting, identification of diseased organs on images, etc. The advantage of AI models is capability of self-learning and improvement in terms of accuracy and effectiveness over time. Their performance depends on the amount of information that they can ingest. It may come from sensors, internet, social networks, databases of special services. They can process photos, audio and video content as well as texts.

The advertising industry has use cases of AI-powered analysis of the faces of buyers caught in the camera: their age and emotional state. As a result of the analysis, the system shows the most relevant advertising in stores. It is also possible to create digital advertising posters that analyze the emotional response of people, and change stories, fonts, and drawings of on the poster depending on the specific person.

- Why do companies and organizations need such technologies? What problems can they help to solve?
- Our economy undergoes the digital transformation, and that is why it has become necessary to analyze large amount of data. You can browse the social networks, look at certain parameters, make some conclusions, but when your task is to study 1 million profiles on Facebook and select the pages most relevant for a specific marketing campaign, human resources are not enough.

AI systems eliminate human errors caused by fatigue, the attitude of workers, or low labor performance. With their help, you can analyze a huge amount of information with incredible speed at a breakthrough quality level — and, in turn, this will help companies to save on human resources.

The main goal of AI is to synthesize human experience. Let me give an example from a recent dialog with the director of a cattle-breeding company: just as an experienced pig farmer visually determines the weight of a pig, a special video recognition system and a camera aimed at the animal can measure its dimensions, and the machine learning and predictive analytics tools calculate its weight.

When solving specific tasks, AI can perform better and faster than people do, and yield better results. That is no secret that computers have beaten people in chess, go, and checkers a long time ago. Thanks to the continuous learning process, a machine can start to solve specific tasks better than a professional in a particular field. At least, it does not require several years to study a subject matter in detail, and only then start making on-the-fly decisions.

All AI models are built to solve one highly specialized task. The advantage of human intelligence is that it is able to handle a variety of different problems, follow a comprehensive approach to problem resolution, think outside the box and be creative.

- But is there something that machines still cannot handle? What tasks still can be done by natural human intelligence only?
- Every AI model is capable of solving only one highly specialized task: recognizing human voices or faces, driving a car in autonomous mode, selecting the most relevant goods for users based on their previous orders, etc. The advantage of human intelligence is that it is able to handle a variety of different problems, follow a comprehensive approach to problem resolution, think outside the box and be creative. You may ask why it is impossible to combine many modules, each them responsible for specific task? Because the science still fails to clearly understand how to synchronize them all. Fine-tuning of these modules, their correct use and improvement — these are the things that cannot be done without human intervention.

In fact, people as a resource will be partially displaced from certain industries in the same way as assembly line production was displaced some time ago. The dangers of intellectual inequality and changes in the labor market are real. Nevertheless, new jobs and new professions will appear too: for example, an artificial intelligence architect, a cyber-criminologist, or a specialist in robotics. According to McKinsey, AI will replace professions requiring physical labor and information processing: retail, hotel staff, etc. There are specialties in which robots cannot completely replace people at the current state of the art: doctors, psychologists, lawyers, judges, politicians, athletes, engineers, and others.

- What are the prospects of AI development in the future?
- The economy will be moving towards the maximum use of artificial intelligence. This will be one of the key technologies underlying the fourth industrial revolution, because they increase the labor productivity, stimulate consumption growth, reduce costs, create new revenue channels, increase the number of customers, and improve the quality of service.

It is no secret that in the near future it will be possible to automate plants by almost 98%. The marketing industry already has an example of a complete transition to an AI-powered business model. One American company suffered from the low efficiency of the advertising agency, with which it collaborated on marketing activities. They decided to transfer these responsibilities to a DMP platform. The artificial intelligence solution has managed to increase the return on investment in marketing campaigns by more than three times and increase the customer base by 30%.

AI will mainly undertake routine processes. In HR it will sort out CVs, contact candidates, conduct initial interviews. In medicine, it will analyze the functional state of patients, study the CT and MRI data, make an accurate diagnosis, and detect particular diseases in the early stages.

In the future, AI will enable a redistribution of tasks between machines and people but, as I have already said, creative and sophisticated activities will still be done by people. Do not to be afraid of robots: they are not our enemies, but our partners. A technological breakthrough in AI is our only hope to deal with economic recessions and crises.

- What AI solutions does Softline offer?
- Softline creates and implements mathematical models powered by machine learning.

In finances we help in credit scoring of suppliers and buyers, configure the systems for determining the possible supplier credit limit similar to that used in banks. For production enterprises we implement projects on process optimization and product quality management. With the help of AI, Softline can help HR specialists rank candidates by CV, and determine the likelihood that a specific employee leaves the company.

For call centers and cold sales, it is possible to rank the contacts in your bases to increase conversion. In marketing, we can implement solutions for client clustering and segmentation, marketing campaign efficiency evaluation, A/B testing, and determining the likelihood of customer loss. Our company can also use AI in such applications as predictive repair and analysis, inventory management, identifying suspicious transactions, etc.

Cutting-edge AI technologies enable the successful implementation of such projects and achieve revolutionary business results.

Want to learn more?
Ask Stanislav Voronin: Stanislav.Voronin@softline.com

Подписаться на новости

На указанный адрес отправлено письмо с подтверждением.
Stanislav Voronin
Stanislav Voroninhead of the BI Department at Softline

Рейтинг материала