The community founded by Darina helps companies solve their problems in the area of artificial intelligence and data science. AI Booster conducted the DS Competition for Vodafone, which solved two problems: determining the subscribers' age and the probability of installing a mobile app by a particular subscriber based on available anonymous data.

AI Booster also conducted several hackathons to solve retail, medical and security challenges and established two technical conferences: Computer VISIONers Conference and Production AI (on the technical aspects of AI implementation). Recently, Daria has entered the rating of 25 entrepreneurs under 25 from

How did you start working with artificial intelligence? Was there a magical moment when you opened your eyes in the morning and thought "OK, I'm doing AI"!

No, everything went step by step; it was natural. System engineering in KPI (Kyiv Polytechnic Institute), work with data, webinars for data workers, community growth, work with companies, and after that, there were real events with applied master classes and hackathons with the participation of leading companies.

Can you name three favourite case studies on machine learning?

The first one is the Hackathon from Vodafone. This company has its own analytical department which determines the subscriber's age using customer behaviour data such as call and migration statistics, and then this information is used to send more relevant advertisements for this age group.

The essence of the hackathon was to come up with an algorithm that works better than the Vodafone analytics department. Oddly enough, but most hackathon teams were able to do it, and the winners showed outstanding results.

The second case is an example of small optimisation tasks that make life a bit more convenient.  For example, buying fruits and vegetables in the supermarket, we always need to memorise their code, then type it on scales to print a barcode that will allow us to pay at the cash desk. There is a development — a camera connected to the neural network, which automatically recognises fruits and vegetables on the scales. And you do not need to remember extra numbers anymore.

The third case is a large number of medical applications that help doctors save lives. Lots of them are now actively being integrated with smart-bracelets measuring pulse. According to them, it is possible to diagnose a future heart attack one month before it happens.

Many Ukrainian teams work in this direction:

  • MAWI is a bracelet that diagnoses cardiovascular disease.
  • Cardioma is a bracelet that you can give your elderly parents, and always be aware of their state of health or promptly call the ambulance if needed.
  • EMwatch is a miracle development that monitors pulse, pressure and cortisol levels. With these indicators and a bit of magic, the device controls your performance, stress level and allows you to optimise the balance of loads.

There is also a range of medical B2B solutions that serve as doctor advisors. Those programs can automatically read the medical files of all hospital patients and predict the main health risks for each patient. Or to recognise various tumours and diseases in X-rays. The complete automatic diagnosis is still far away because it is primarily a matter of responsibility. Because if the doctor misses the diagnosis, he is responsible for it. And if the algorithm is wrong, then who will do it?

How many professions face automation?

I do not think that the algorithms will substantially crowd out people from the labour market, although there are more than enough apocalyptic predictions on this topic. MIT predicts that automation will take 80% of jobs over the next five years. Oxford Martin school forecasts a loss of 50% over the next ten years.

Still, there is a large number of creative professions. Despite the fact, that the algorithms can already create music, and draw paintings, there is an abyss between the level of people's and machines' content.

The work of recruiters for simple professions, for example, is well automated through bots. There are pilot projects that can read the resume. Companies are now actively automating technical support services using semantic analysis algorithms. For example, if you write to banking's support, then the special algorithm first analyses your message and tries to suggest solutions based on the keywords. Much of the routine conversions are now processed automatically.

What do you think about the growing number of algorithms that pass the Turing test? Is this a sign that universal AI is approaching?

The Turing test is an outdated metric. Today, there are chatbots, which you can confuse with a human-being — it is no longer a supertask. There were cases when the Turing test was passed solely due to developers' creativity: for example, the chatbot made grammatical errors, and the test commission thought that the robot would write correctly. In general, a universal AI is not our topic. Enormous corporations and state labs with tremendous funding develop it, and everyone marks these developments top secret. We are dealing with more useful things.

How do you assess the impact of artificial intelligence on a cyber sport? Now there are matches like "top players against AI" almost on every major championship and even games among different AIs.

I am positive about this. Artificial intelligence in games significantly promotes the very topic of AI development. Also, many startups develop AI assistants to cyber-sportsmen. For example, 20thousandleagues is an artificial intelligence that teaches you to play the League of Legends. Unlike the usual interactive tutorials that teach game mechanics, this one explains a team game and even the right communication with other players.

Robots that teach people to be better. What a surprise! What other science-fiction stuff has already become a reality?

Quite a lot. For example, "Hollywood zoom" is a technique from detective films and science fiction, when an image is magnified, and its quality does not drop, and pixels do not appear.  Now there are neural networks that recognise the object in the picture, and add new pixels with magnification. It is a reality. Or Ukrainian development Reflect: you can put another person's face directly into the video. By the way, it was created by the same team that produced the famous Prism app. Now this face change on the video is still noticeable, but technology is moving towards improving it.

This technology plus the PornHub video base... Is there anything out there that is beyond the reach of artificial intelligence?

Yes, there are a lot of things — for example, simple voice communication. Lots of clients want us to create a bot that will perform a live conversation with customers. But if you've ever talked to Siri or Alexa, then you know that voice control technologies are not well developed today. Their corporations have hundreds of smart programmers and massive databases, but they can not make an artificial intelligence that naturally supports conversation and recognises commands well. Robot Sophia is an example of a fake AI. It has a built-in microphone and camera, and the person behind the microphone gives answers instead of the robot. Also, in the field of image recognition and their categorisation, AI has not gone so far.

How significant is the progress of hardware in the development of AI?

It plays a crucial role. Neural networks existed in the '70s. But only now they become commercially available, and all because of the significant fall in the prices of information storage (and its transmission over the Internet), as well as teraflops of computing power. These are quantitative changes that turn into qualitative ones, and allow you to do things that we previously could not imagine.

What would you wish our readers?

Pay attention to databases and streams. If you have ideas on how to use them to do something better, contact us.

WTFBit Media writes about blockchain and emerging technologies in a simple way, with humor. We also work with diverse clients helping them effectively deliver the message to their audience.