The artificial intelligence market is growing. While society is debating the ethical aspects of the digitalisation of everything, business is already making a profit out of smart digital helpers. What sectors are already using artificial intelligence? What are the advantages and shortcomings of digitalisation? Will increasing use of AI eventually lead to mass layoffs? What high-tech solutions are accessible to small companies? In the interview, Arb Pro Training Institute Group’s head of strategic planning Roman Koposov told Kommersant UK about all these things.
What is the size and state of the AI market?
In 2019, before the effect of the pandemic, the world market for artificial intelligence stood at $36.5 billion (£32 billion). That year it grew by 30%. According to the figures of leading analysts and developers, this trend will continue until 2023. We evaluate this to be a conservative forecast, and that a growth rate of 40-45% should be expected. Here, we don’t just mean investments in out-of-the-box products, but also in fully integrated AI solutions. There are about three or four different types of solutions which most people regard as artificial intelligence. The first of these is machine learning. This includes predictive analysis and smart data analysis. The second class is everything to do with computer vision, which is when algorithms process images, either live, from cameras, or from pre-existing images and patterns. Using machine vision systems, handwritten text, road signs and people can all be recognised. These systems are vital elements of self-driving systems for cars, they allow handwritten texts to be deciphered and they can also establish people’s identities based on their appearance. The third class of solution is the development of natural language processing and speech technology, examples include voice bots such as Alisa, Alexa and Siri amongst others. These attempt to conduct conscious dialogue when talking to people. All of these classes of solutions are present both in Russia and in Britain.
In what areas is AI most often used?
The first area is an advertising and the media. Due to the initial development of the internet, email services such as Google, Yandex and Mail.ru possess huge amounts of data. Together with the telecoms sector, they were the first to start to integrate various solutions based on artificial intelligence to tweak search results and advertising to ensure its relevance to the majority of searches. The second area is financial services. Credit scoring is one of its main tasks. Major financial companies in Russia and abroad calculate this using AI. Algorithms based on AI allow people’s so-called digital trails to be traced (including social media profiles). This information can be used to determine people’s reliability and whether their credit ratings need to be corrected. I think that in the future, there won’t be a one size fits all approach to interest rates on loans. Instead, there will be an individual approach to each borrower. Other large sectors using AI are insurance, retail, logistics, supply chain management, manufacturing and agriculture. Law firms round off the list.
What are the advantages and drawbacks of introducing AI?
Let’s look at the advantages first. The first of these is economic; they improve efficiency and reduce costs as they free up human resources. The second is risk management. There are two advantages here: the boost to profits a company can receive by using sensors to collect information allowing it to optimise its working practices, and the improvements to checks and security measures which can be made when cameras are installed to observe the manufacturing process, monitor staff behaviour, see how personal protection equipment is worn (for example, masks in food production) and whether safety guidelines are followed. The third advantage is the heightened speed of decision taking (in retail, AI quickly analyses consumer preferences in order to send each client personal recommendations). It can also identify new priorities for consumers. For example, if someone has stopped eating fatty foods and started leading a healthy lifestyle, then their choice of groceries will change as a consequence. AI analyses their recent purchases and suggests appropriate food products, and the client feels valued.
As for the drawbacks, let’s disabuse ourselves of the fanciful notion that AI will be able to take over the world. Storylines like this feature quite often in films, but in my view, these fears are completely unfounded. Alone, AI cannot create anything. It must be fed data which a human gives to it to perform a task. The main problem now when solving business tasks is excessive expectations. These arise primarily because people don’t yet fully understand the capabilities of artificial intelligence. It goes without saying that AI can solve a great many problems, but to do so it needs consistent data to work on. No accurate calculations or conclusions can be made with ‘dirty’ data.
At the moment, many people are saying that the increased application of AI will lead to mass layoffs and that these will particularly affect call centre staff. How well-founded are these concerns?
So far, it is too early to say that AI will replace a large number of call centres; they currently solve complex and contextual client requests. It will take AI a long time to learn to do this. So there won’t be any mass redundancies for the time being; they will happen gradually. And let’s be honest; the most highly qualified professions will not be replaced, what will go is repetitive human labour where people work according to scripts (such as an operator who sells using specified phrases, or an insurance consultant entering handwritten data). It’s the same with the robotisation of production, where only workers performing simple operations will be replaced, such as placing goods on pallets. Companies such as Danone, Nestle and Mars are watching these developments with interest, as they should mean that seasonal workers, who are currently very expensive, will not be required in the future. What’s more, robots and AI don’t need light, motivation systems or wages; the list of advantages like these is endless. The main thing is that they give predictable results, and even if mistakes can still occur, then these come to light and they are not repeated. More and more automated checkouts will appear, and self-service shops like Amazon Go. The company has already opened around 27 small outlets where you can buy all the essentials, pay and leave. Of course, the customer may take the wrong goods, but if this happens, then the machine vision system determines what has been taken in error and in what quantities, identifies the customer and solves the issue. So low-qualified retail staff will be laid off, but it won’t be the same for everything related to creative activities and complex manipulations (such as sorting groceries into different categories and placing them onto shop shelves). It will be a long time before these tasks can be replaced.
In logistics, robotisation and machine vision systems work as assistants. For example, on the basis of a driver’s pulse and biorhythms, these systems can warn them that they are tired and recommend they take a break and relax. They could also inform a fleet manager that a specific driver is violating traffic regulations, or has a fuel-inefficient driving style, or that they brake too hard, which quickly wears out; tyres, or recommend that the driver be sent for a road safety training course.
In medicine, the introduction of AI will not lead to any redundancies, as the technology has a predominantly advisory role. Firstly, electronic data gathering allows AI to analyse the full picture of a disease, by comparing it to a great many cases in other patients, sometimes millions of them, to determine which treatments were most effective in similar situations and decide what to recommend to the patient. This will not replace the doctor’s recommendations, but it provides clear criteria which need to be taken into account. Insurance companies working with health institutions like these systems most of all, as they give protection from doctors’ errors. All countries are anxious to get doctors’ training right, but even the best-trained physician might miss something. These systems offer protection against this danger. Secondly, AI-based systems in hospitals can warn medics when a patient’s condition worsens, or warn of changes that may be danger signals. Or, if the ward is equipped with cameras connected to a machine vision system, they can simply predict when a patient is at risk of falling out of bed. Companies that develop these systems already exist. Some of them are Russian. They produce solutions for clinics in the US and Britain. They’re in enormous demand.
What’s the difference between AI and a robot?
The task of robots is to liberate people from mechanical routine tasks (such as welding, painting and logistics). They do it accurately, with minimum wastage, and expend fewer resources than humans. AI works with data (images, voice, figures and other forms of data. There is also the subdivision of Robotic Process Automation. This is the automation of processes using programmed robots, which, over the course of several weeks, analyse an accountant’s work day, to take one example. They identify the unnecessary operations which are repeated most often and can be cut, and write an algorithm setting out how to do this. This gives the human more time to perform more important managerial tasks.
Can AI replace any professions in medicine?
In theory, yes. For example, there are some endocrinological disorders which do not require the physical presence of a healthcare worker to be treated. This is why telemedicine has developed. This technology can quickly analyse what someone is saying and give recommendations. AI can replace a doctor, analyse test results, compare symptoms and various factors such as the patient’s workplace or place of residence etc. However, the question arises of what responsibility the creators of such technology would have. Right now, no one can answer this question from a legal perspective. What would happen if someone died as a result of the AI’s recommendations? So far, this is only a topic for science fiction programmes. This is one issue. A second one is that doctors’ skills and knowledge might begin to gradually deteriorate if they know that all their work is being checked by AI. This is why it’s important for an experienced physician to give their version initially, and for it to be then checked by AI for confirmation. That should work well. For example, there is a company called NEALBE which develops and produces innovative devices for monitoring the body’s vital signs. They have created the GoBe bracelet, which automatically tracks the absorption of calories from food, water balance and stress levels. As well as highly accurate sensors, they also have a support service which allows the software to be customised to the individual’s lifestyle and eating habits. Additionally, the company works with telemedicine and gastroenterologists who can give advice on what steps to take based on the data provided by the bracelet’s AI. Many great things can happen when the algorithm interacts with the AI, and this is helpful for doctors. This is because no doctor can guarantee an outcome, so patients always pay them for a process. Here we have an out-of-the-box solution which cannot be tricked because the bracelet counts the calories you use and you have absorbed. The company supplies its products to Britain.
What AI solutions are available for disabled people and those with severe difficulties communicating?
A huge breakthrough has been made in this area which allows people in vegetative states to communicate with the world, which is extremely important. First of all, there are neural interfaces which allow signals from the brain to be interpreted as speech or text and help people to interact. Secondly, even Russian clinics are now using exoskeletons for people’s recovery after difficult operations, when the musculoskeletal system is damaged. Many have got back on their feet thanks to this technology. This is predominantly robotics, however great computing power is also required. A third development is bionic prosthetic arms which require specialised programming solutions involving huge volumes of calculations. The most difficult part is finding investors for these projects, promoting them and making all efforts to ensure they reach the market.
Is AI used to solve any psychological problems?
We’re talking about assumptions. Now we must return to the voice assistants we talked about earlier. They are getting smarter and more and more difficult to distinguish from the voices for real people, despite nuances such as voice timbre, pauses, etc. This means it is possible to use AI, but it is unlikely to replace experienced psychotherapists, psychologists or coaches, especially at first, as they work with personalities with different aims and results. Here we are touching on an important legal point, since all patient histories, as a rule, are only accessible to the specialists working with them, and they have no right to place them on the web or anywhere else. Theoretically, they can be anonymised as much as possible, but then the question of allround data security arises. There are already startups working on this such as Lyssn.io, which applies NPL technology in the service sector. The British service Leso has been using methods based on big data (primarily NPL technology) to treat depression and nervous disorders since 2013.
How accurate is the claim that AI is already in widespread use to watch people?
Russia and China top the list of countries using СCTV surveillance cameras for face recognition. For the average person, this technology became noticeable when football fans weren’t able to buy tickets to matches because the system had recognised their faces, checked their criminal records, and identified them as troublemakers. During Covid, inhabitants of Moscow who were supposed to be in lockdown who left their homes later received fines for doing so because cameras had recognised their faces and identified their full names from their IDs. Why does Russia stand out? In this country, there is a company called NtechLab which has developed a face recognition algorithm that is better than those developed by many other companies. It has won international competitions for its speed and accuracy. This is a fundamentally important point. Their systems have been used and are being used in face recognition cameras. This is a general point. If we look at open access sources, we see that airports such as Sheremetevo and Domodedovo say that they are using cameras to detect behaviour that is atypical for ordinary travellers, in the search for those with a propensity to terrorist attacks or theft etc. Let’s get back to retail…In theory, there is a machine vision system allowing a thief to be identified at the moment of their crime by following their movements. I am 100% certain that, at customs, machine vision systems are already in operation to recognise faces and compare them with photos from social media in a search for people who have something to hide. Before approving a loan or mortgage, banks could take photos of potential clients and run them through a social media and digital trail check.
Is it possible to create AI which can learn independently, like a person does?
An honest answer is that there is still no clear definition of AI. It is a pop science concept, like emotional intelligence. It’s a term which is currently in vogue. AI algorithms feed on what they’re given. Comprehensive AI is a slightly futuristic topic. It takes the form of a hypothetical robot which can see, hear, generate speech, move from place to place, perform manipulations and constantly teach itself. AI does not make the same mistake twice, that’s not the way it’s programmed. But it can only perfect itself using the database given to it. Creativity and making something out of nothing are not functions of artificial intelligence.
What British companies are advanced in their application of AI?
The top companies for AI coming out of England are, in my view: the independent geopolitical analytical and consulting company Oxford Analytica, which relies on a global network of experts to advise clients about investments, strategies, politics and operations, OneTrust with its Trust Intelligence Platform, which combines data, teams and processes for trusted cooperation including in the cultural sphere and Wayve, with its AV2.0 technology which completely rethinks how to solve the problem of autonomous driving by taking an approach entirely founded on machine learning. The result is the technology of the next generation of AI, which can be scaled up and adapted to any driving conditions in any place on earth.
What corporations have achieved the greatest successes with AI, and whose use of AI has ended in fiasco?
AI is still regarded as a little risky; it seems that it’ll be crucial in the future, but it requires enormous investment. Almost all IT and Fortune Global 500 companies use it. They are in the AI ecosystem either thanks to their hardware or their software. These include Apple, Google, IBM etc. However, number one, in my opinion, is Tesla, which is currently making maximum use of AI. We’re talking about the economies, high-speed decision making and risk management provided by the integration of any type of AI. Tesla has created a technology with new value for consumers with its new self-driving system.
I suppose that in the foreseeable future, an enormously groundbreaking new IT company won’t emerge, as everyone is afraid of missing out on a potential new market. Back in the day, Microsoft lost the browser and search engine market because Google appeared, which, in its turn, was not able to dominate the social media market (failure awaited Google Plus). Facebook was unable to take advantage of a new trend in social media and create a TikTok analogue. Now they’re making a bet on VR, as currently everyone is investing huge funds into all conceivable aspects of this technology and buying up startups. Some are doing this to inflate their capitalisation and make themselves more attractive to investors because many of these are buying startups and failing to integrate them into their companies. Others are doing it to take large-scale or unique technology and grow it even more quickly so that a small startup can become part of a large company.
As for failures, I have heard of virtually none, as no one likes to talk about them. The American company Ring caused a scandal when they developed a security device (a smart doorbell with a camera), which warned people if an intruder entered the house or if something was broken. According to the Intercept, the neural network, which was supposed to recognise children and tell a normal situation from an emergency, was developed by a team of programmers from Ukraine. So a load of data about how American taxpayers spend their time ended up in another country, although it wasn’t leaked to another neural network. Later, Ring announced that their staff had never given third parties access to live video streams from their devices. In fact, they hadn’t even had access to these streams themselves. Nevertheless, there was severe damage to their reputation.
What forms of AI are available to small and medium-sized businesses which do not have huge sums for investments?
These technologies range from bespoke, high-cost solutions made to individual orders, to mass-market, out-of-the-box solutions capable of solving problems with sales, human resources and production. They are affordable for relatively small companies. For example, solutions in the area of control can cost less than one million roubles (about £14,000), while an advanced chat robot with elements of self-learning will set you back around 500-600 thousand roubles (about £7,000-£8,500).
