NORBERT BIEDRZYCKI
9 min readPressRead this in Polish

The soulless logic of algorithms seduces business and flirts with us, the customers

The soulless logic of algorithms seduces business and flirts with us, the customers

I will present, in the most condensed form possible, the technological foundation on which phenomena such as autonomous vehicles, recommendations during online shopping and the spoken answers of mobile assistants are born today. Will the logic of algorithms kill human emotions?

My article in Spider’s Web – link to the article (in Polish).

The machine learns without end

The two basic forces that determine the key processes of intelligent technologies are machine learning and deep learning.

Machine learning is often defined as strong support for predictive analytics. It is a technique from computer science and statistical modelling that allows a machine (an application), through its own analysis and without being explicitly programmed in advance, to predict the outcome of an action and often to work out and make a decision on its own.

Deep learning, in turn, is a type of artificial intelligence based on neural networks, whose main task is to improve voice recognition and natural language processing. Simplifying a great deal, it can be treated as a way of automating predictive analysis.

In short, these are techniques that make it possible to process enormous data sets through machine self-learning. They differ in their capabilities and tasks. For more on these techniques, see my previous articles.

The algorithms behind machine learning work on a scheme in which the computer takes input signals (data) and decides what to do with them next – how to group them, classify them and so on. Machine learning has an important limitation: when performing tasks, the computer works within supervised experience.

This means that a person still plays a key role, “feeding” the device portions of information. The world’s most powerful computer, IBM Watson, could not achieve masterful results if people had not first “fed” it data or pointed it to sources of data and knowledge.

The second technique, deep learning, which is responsible for voice recognition, natural language processing and image recognition, is a qualitatively different level. It resembles the operations that take place in the human brain. The computer uses neural networks, made up of groups of processors working in parallel and fed with large amounts of data.

After each new experience, the connections between these processors and the streams of information they handle – the nodes, to simplify greatly – reorganise themselves. These organisations become ever more purposeful, so the machine responds more efficiently to the tasks set for it.

I admit I am not immune to a sense of something uncanny about these seemingly purely technical processes. Machines drawing on the history of their own achievements and gaining experience through contact with people never cease to amaze me.

How do you recognise a dog in a photo?

The fact that computers can solve complex problems ever more efficiently, thanks to work they have already done and contact with people, is producing a breakthrough in the development of intelligent technologies. Fine, but how does all this translate into the market popularity of particular technology trends? Once again, there is a certain simplicity here.

Machine learning and deep learning analyse data in a few basic ways, formulating a set of decision rules on their own or with human help. The most important are: database analysis (the basis of big data), classification of objects (organising photos by shared features), error detection (comparing objects and states to detect the likelihood of anomalies), recommendations (the basis for purchase suggestions) and optimisation (finding the best route for a car).

Everything that matters so much to consumers today, and that drives business innovation, is based on these few operations, presented in an attractive package.

Norbert Biedrzycki blog autonomous vehicles

A truly divine purposefulness

Under the influence of business, marketing and social decisions, the processes described above turn from IT abstractions into concrete innovations. If algorithms processing data can “predict”, the logical consequence of this ability is e-commerce applications in which a shop offers customers goods similar to things they bought before.

Another example: if algorithms can classify data by shared features, they will be able to recognise flawlessly whether a photo shows a dog or an aeroplane. And that cannot escape the attention of the owner of a social network that collects billions of users’ photos. Within artificial intelligence, then, we are dealing with a certain logic and purposefulness.

Below I look at three trends that, thanks to the mechanisms described above, will consolidate their popularity in the world of digital business.

Algorithms can eliminate errors…

The error-detection function mentioned earlier will become increasingly crucial, given both public opinion and the popularity of certain business trends. Controlling and eliminating anomalies is particularly important in the Internet of Things. I mean the growing popularity of devices that use sensors to collect data from their surroundings.

The huge amount of information collected from our streets, shops and even homes (smart houses) brings indisputable benefits. But it also carries danger. For the average citizen, irregularities can mean both everyday nuisances (smart devices breaking down) and a sense of threat (harmful monitoring used for political purposes). For business, the matter is just as important.

Consider safety, which is crucial in the automotive and aviation industries. Sensors used in aircraft engines process streams of information generated in real time during flights. Eliminating potential errors in these processes is enormously important. Sensor systems also give autonomous cars their abilities. Eliminating irregularities in data collected by cameras and lidar sensors has a huge impact on the further development of the trend.

… know how to satisfy the consumer…

Using the predictive functions described earlier (recommendations) is a priority in e-commerce and retail. Companies such as Amazon and Netflix know very well how valuable automatic recommendations are. They turned predictive systems into a global business value and in effect created a new standard of behaviour for the average customer.

In the coming years, e-commerce will see the development of technologies focused on analysing customers’ voices and facial expressions. The analyses I have in mind will be used to train bots – digital assistants able to communicate with us ever better in natural language.

Today’s still imperfect bots will steadily improve their skills until, in time, they fully compete with call-centre staff. They will suggest products we are interested in, book our holidays, point us to the best loan and make our doctor’s appointments.

… and understand the need for efficiency in business

The predictive skills of algorithms can have an enormous impact on an important business category: efficiency. Increasing it is, as we know, a priority for many boards, whatever the industry. Any technology that shows we can make better use of existing resources naturally attracts business interest.

It is precisely this ability of algorithms that makes it possible to reduce costs and increase productivity. Take transport, where algorithms can bring a great many benefits. Heads of transport companies will not be indifferent to the fact that real-time observation of drivers and vehicles can cut costs by as much as a dozen or more percent.

Devices that track the driver’s reactions, engine performance, stops and obstacles on the road can suggest when the driver should speed up and when slow down, when to refuel and when to keep going. Every decision that takes such information into account translates into financial savings.

Norbert Biedrzycki big data AI congress

Who can already appreciate all this?

In its report “Notes from the AI frontier: Applications and value of deep learning”, the McKinsey Global Institute analyses which industries may be most interested in deploying these technologies and which can turn prediction, image recognition and error detection into business value. Retail currently tops the list.

Its leading position is no surprise, given the huge amount of customer data it has today. Information about age, education, work and previous purchasing decisions lets retailers take the customer experience to another level. Combining what a customer’s purchase history says about them with the content they generate on social media is becoming a standard trend in global e-commerce.

The logistics industry mentioned earlier is also among the industries that value artificial intelligence tools most. The high position of the travel industry and its commitment to new techniques can be explained by several factors. For a travel agency, tracking a customer’s travel history and analysing the photos they share online are now essential to stay in touch, offer new deals and so keep the business profitable.

The huge popularity of geolocation and travellers’ activity on social media are two factors driving significant growth in available data, which is a tempting morsel for the industry.

These examples show that every industry in fact uses a mix of different algorithmic functions. Recommendation systems, error detection, anomaly detection and planning the best decisions – only in combination can they create the technological base that decides a business’s success.

Obstacles we must overcome

The McKinsey Global Institute points out that of all the companies surveyed that said they were aware of the great potential of artificial intelligence, only 20 percent had taken steps to adopt it. Several inhibiting factors still mean that, in business practice, not all the possibilities of intelligent technologies can be appreciated and adopted straight away.

I will start with cognitive discomfort, which can be summed up in the question: “Why is the technology suggesting this particular behaviour to me? How did it get there?” The speed at which processors grow more capable does not automatically translate into greater human trust. And this is not only about the irrational fears of ordinary citizens.

The systems of regulation and certification in many industries today are so extensive that rapid acceptance of even the most innovative solutions is not possible everywhere. While marketing will have no trouble with rapid deployments, medicine, aviation and banking will not abandon formalised control procedures and certification systems overnight.

Another limitation concerns security. We are experiencing a crisis in how we think about data processing. It is not easy today to convince everyone that facial recognition systems do no harm. All “data processing” is under close watch, so artificial intelligence, which feeds on our data, may feel the fears of the human species “first-hand”.

Money also matters, of course. A decision that a company should invest in expensive IT systems, create departments dealing with new data and hire specialists must have a rational, economic justification. And as we know, spending money quickly is not a favourite activity of boards, even at the most progressive companies.

AI Conference TIME Norbert Biedrzycki

Matching value to needs

I think the key challenge for today’s boards will be to recognise and name all the value a new technology can bring. It is not about constant admiration for the global possibilities of artificial intelligence. What counts is matching its possibilities to the specific needs of a given company. The phrase “artificial intelligence” alone no longer opens every door automatically.

Representatives of start-ups and technology companies must understand the specifics of the client’s industry if they want a chance to present their offer. Still, I am convinced that new technologies’ share in market change so far remains modest. That is not a pessimistic observation at all. On the contrary, the market remains wide open to new possibilities.

My article in Spider’s Web – link to the article (in Polish).

Related articles

– Who will win and who will lose in the digital revolution?

– Artificial intelligence is the new electricity

– Robots are waiting for judges

– Only God can count as fast: the world of quantum computers

– Machine learning: computers are no longer infants

Read next

  1. Data Driven Investor: Time We Talked to Our Machines

    My article in Data Driven Investor published 12th of September 2019. We talk to our machines more and more often: the smartphones, computers and digital assistants on our desks. Are keyboards, and the manual operation of electronic devices, becoming obsolete? Over the last few years, artificial intelligence (AI) has

  2. Military robots may decide whether we live or die

  3. BUSINESS INSIDER: How machines think

    My article in Business Insider published 22 October 2016. Technological advances make you wonder: is this the start of a new world ruled by intelligent machines? Is there such thing as artificial intelligence? The development of artificial intelligence in robotics is not as unlikely as you might think. A