AI Conference 

How do e-commerce, banks and insurance companies get rid of crashes and fails caused by the high IT support load? Imagine a big e-commerce case: stable model predicts a change in the IT resources load depending on the growth/decline of business indicators and various factors (stocks, sales, etc.) and also increases business results.

We’ll talk about how, chatbots have turned into smart assistants and learned to solve real business problems in recent years: from accepting applications to automating internal processes. We will discuss the integration of chatbots into the business environment: types of tasks for bots, ways to find business processes that require automation, as well as which bots will be in demand in the next 3 years.

The presentation will provide real case studies in the sphere of biometric identification introduction as well as possible difficulties, needed resources and what to expect from the technology adoption in the future. We will also talk about the automation of identity identification in existing service channels as well as the inclusion of remote client identification in remote channels.

While many people are striving for the full replacement of human being in certain areas, let’s step back and examine the augmented intelligence concept. The speaker will reveal how IBM helps to create assistants to various specialists as well as what pitfalls one should expect, and what it can lead to.

The following issues will be covered:

• Solutions to key business tasks using the example of a platform for developing smart chatbots: how to quickly create a bot, debug its functions, train it to understand the natural language.
• Creating behavior pattern of the bot.
• Cases of introducing chatbots in various industries.

Deep neural networks are a core of today’s artificial intelligence. Nowadays, they allow us to successfully solve problems in such areas as computer vision, natural language processing, and mining of other data types. Moreover, the technology of deep neural networks learning proved feasible for creative tasks resolution: a synthesis of unique content from some concept or idea. The presentation will cover the mechanism of today’s neural networks working and learning, the nature of deep learning and its application spheres.

The medical image analysis solves a whole range of machine learning tasks, including segmentation, classification, search and synthesis of medical images. Several key features of medical image data extract significantly limit the use of conventional deep learning methods developed to analyze two-dimensional images. First off, data high dimension is accompanied by a quite low volume of learning sample, while the application of data augmentation methods is pretty limited. Besides, a big size of 3D images imposes certain restrictions and leads to the development of specific convolution network architectures. Finally, data tool variability has a significant impact on the analysis outcome. During the presentation, we will discuss the use of deep learning methods to solve such tasks and provide several examples of neurovisualization data analysis projects.

Mobile Messenger Marketing as a necessary part of companies’ digital strategies in 2018–2019. Messengers are going to become a primary entry point to the Internet, development prospects in Russia and threat for Yandex and Google search systems.

Why large companies have already noticed chatbots.

How do they help save money, increase sales and customer satisfaction index.

Successful business cases, metrics, development plans.

Deep learning of neural networks has been a breakthrough in the evolution of artificial intelligence systems. Many tasks lying beyond the reach of classic machine learning techniques have been solved at the level close to human abilities or even at a higher level. However, there are significant problems in teaching neural networks that seriously restrict their application. Examples of such problems include a large amount of marked data required to teach a network, difficulties of transferring the experience of a trained network on other tasks. The speaker will elaborate on these problems and possible ways of their solution.

Modern neural networks are based on knowledge of the brain function of the early 20th century. There are fundamental limitations, including the training and stability of these algorithms. There have been many significant discoveries in understanding the brain over the past 20 years, which allowed us form a new concept of the brain function. This concept formed the basis of the development of a new type of neural networks that will work with the meaning of information, create strong generalizations and strong AI.

Computer vision is one of the most important components of AI, it is used to solve a wide range of practical problems: from collecting smart data from surveillance cameras video stream to automatic tag generation in photographs. Traditionally, the introduction of such technologies requires not only high qualification of developers, but also huge amounts of training data - hundreds of thousands of examples and more.

We will look at how computer vision tasks can be implemented using various solutions from Microsoft: starting with solving typical tasks based on cognitive services, using the custom vision service to create recognition models and also on the basis of neural networks using Cognitive Toolkit. We will take a look at the technologies themselves and give some examples of using image recognition in practice.

Since Henry Ford’s invention of assembly line, the efficiency of physical work has not changed crucially. SOLUT develops an industrial IoT project aimed at analyzing the efficiency of staff labor time in the manufacturing segment, construction sector, housing and public utilities, and other branches of economy.

The project goal is to increase the working efficiency of blue-collar job representatives by designing portable devices to determine down time, critical abnormalities of technological process, and violation of safety techniques.

Devices monitor workers’ activity during their shifts using acceleration indicators and gyroscopes as well as recognize types of activities typical for every certain specialty using patterns based on machine learning methods.

• What are the tasks of retail chains?
• What are the tasks of goods manufacturers?
• Why retail chains have not yet implemented shelf space monitoring.
• How FMCG brands have become technology proponents in the retail sector.
• How computer vision changes business processes for the largest FMCG companies.