Leading the trend, part two

Hello everyone! Pavel Krasovsky, Deputy Director of the Center for Strategic Innovations at Rostelecom. I wrote about the importance of my own methodology for assessing promising technologies in a previous post, in short – we need to understand exactly what criteria are used to assess the importance and relevance of a particular trend. Third-party research is based on non-transparent methods, so “If you want to do well, do it yourself.”

Methodology

The cornerstone of our methodology is the obvious fact that a huge amount of information has already accumulated on the path of human development, especially in the digital age. All this, from knowledge about the technologies of industrial revolutions of the last century to the freshest ideas, has already been digitized and carefully stored somewhere.

Moreover, it was important for us that for each stage of the trend development there was a source that we can trust, otherwise it puts an end to the entire methodology. We identified several sources that are self-organizing and self-updating systems, where it is beneficial for authors to post information:

Scientific environment.
Scientists researching a particular field are themselves interested in actively writing scientific articles. Here you can improve your own rating, and the citation index, and the position of significant magazines, and many other goodies.

Inventions and patents
With patents in general, everything is simple: he was the first to patent something worthwhile – he was provided for until retirement. And all information about patents is public, because everyone wants to stake out a clearing.

Creation of startups.
Startups, with rare exceptions, are themselves interested in disclosing information about the amount of funds raised, because it shows capitalization to other investors and helps the startup raise more rounds.

Positions of major industry players.
Companies are interested in spreading hot vacancies in order to close them as soon as possible. Everything is clear with the media – when launching a new product, any company by hook or by crook tries to get into the media.

And some of the whales can take and buy one or another startup. Or go and start making your own startup in the same industry and on the same topic. We also track this both in specialized resources and in news and resources for HR. What does HR have to do with it? When a large company does not buy a startup, but decides to make its own on the desired topic, it begins to build up competencies, read – to hire specialists from the required industry. And here it makes sense to evaluate the bases of vacancies and resumes, the totality of vacancies = demand for technology and trend, the totality of resumes = supply. All of this can be collected, counted, looked at intersections and kinks, and found causal relationships. This is how these sources work for many years, which means that they can be trusted.

In general, it would be ideal to deeply study social networks on this topic, but they do not yet provide the necessary level of access to collect such information. Especially Facebook, after the Cambridge Analytica events.

So far, in three years, we have driven through the system 4.8 million scientific publications, 2.4 million patents, information on investments worth $ 2.3 trillion, 2 million vacancies and 7 million resumes, about 1 million publications in the media and the same number of search queries.

How do we rate

It is clear that no matter how much the AI ​​is trained, no matter how detailed models we build, the operation of the system is the first stage of filtering. Everything that we receive with the help of our robots and apishe is then sent for linguistic assessment, and then machine learning clusters and counts all this.

Trending technology stack:

First, the most frequent and relevant data are extracted from the array, which can already be considered trends. A detailed list of trends is built using the hashtags of scientific articles, which expands for investments and patents, there are also quite a few words for analysis, as a result, about 2-3 thousand trends are obtained. Based on the cluster analysis, we connect the vertices, and the machine begins to suggest that a trend is quite possible here and there. At the exit, there is already a list of about 200 pieces instead of 3 thousand.

We are already looking at these 200 trends manually and discarding some frankly general ones. Let’s say the machine gave us the “Internet” or “software”, we don’t need such a level of abstraction. After viewing the entire list, 100-120 trends remain.

Trending process
Trending process

Trending process

By the way, about linguistic analysis. Everything would be much easier if each term was used only in its own meaning, without synonyms and abbreviations. But, as you know, the situation looks different. Somewhere the Internet of Things will be called the “Internet of Things”, in other sources – “IoT”, in patents – “sensor network”, “machine type communication” and so on. In general, who is in what is much, so we adapt the terminology for a specific type of source: for each, a semantic core that is unique for him is compiled, and the trend is tied to a specific lexicon.

Then we take the obtained semantic kernels and run all the data through the system, obtaining quantitative estimates not only for different trends, but also for industries, countries, organizations.

So far we have studied the telecom industry well (we are from telecom), but this is possible for any industry. By the way, we are slowly starting to provide various companies with access to our product.

Where all this can be used

First, it will definitely come in handy for any analyst and strategist in his current work. Secondly, decision makers love this. A large number of different charts can be built on the basis of all our data.

For example, a high-level list of the “Top 100 Anything” type, which the tops love to look at to understand what to invest in. The practical meaning of such ratings is as follows. The company has a certain budget that it is ready to spend on innovations. And she needs to determine what to invest in. Artificial Intelligence? Blockchain? Self-driving cars? Or maybe quantum computers? Lists like these help you compare the prospects of different technologies with each other. Of course, after building such lists, their verification and analytical interpretation is necessary, but it is quite possible to rely on them.

At the end of 2019, the list of Top 15 Trends looks like this:

For the second year in a row, artificial intelligence takes 1st place in the overall rating (over the past year, the gap from mobile networks has only increased). In 2019, cloud technologies showed good growth, moving up from 10th to 4th place, and augmented reality technologies, rising from 24th to 14th place.

We prepare the following cards for each technology:

Analyzing the information received, you can get a lot of insights. For example, while analyzing data on artificial intelligence and quantum technologies, we found an interesting pattern: since 2015, the term “quantum machine learning” appears in scientific publications (the use of quantum computers to analyze data using machine learning). And in 2019, every 15th scientific paper on quantum technology contained references to artificial intelligence. This suggests that scientists are concerned about the lack of current computing power for the further development of AI and, apparently, a quantum computer will be the solution to this problem.

Over the year, 5G has evolved from an innovation into a mature technology that has a fairly strong impact on other technologies: in the countries with the first commercial 5G networks, patent and investment activity in the field of VR has increased.

This year, we noticed that the trend for self-driving cars has grown the most in all five years of observation. Investments in drones have been kept at a high level for two years in a row, which means that the technology is ripe enough for mass use. It remains to address the lack of necessary infrastructure and regulatory constraints.

Another discovery for us was persuasive techniques – a mixture of familiar IT services and psychological techniques. In 2019, there were several large investments at once and the growth of vacancies in this area. The main applications are applications dedicated to a healthy lifestyle and educational services. Another application is the use of these technologies in election campaigns, a prime example: the infamous Cambridge Analytica in 2018. It seems that among the dual-use technologies has arrived.

Trend monitoring is a great tool to monitor the technological development of countries. For example, this is how the struggle between the United States and China for world leadership looks like, which China has systematically captured over the past five years:

China is the absolute world leader in patents and scientific publications: every second patent and every fourth scientific article in the field of ICT are Chinese. The US leadership remains only in the area of ​​investment.

For another reason, tracking trends is important: do you remember how quickly the Apple Watch burst into the market, becoming the world’s first brand in watch recognition in almost two years, overtaking the classic digital brands? It seems like it happened quite quickly and suddenly. In fact, Apple began actively patenting technology for them ten years before the first watch was released.

Therefore, you need to track technologies in the early stages, when they are just emerging. We have also learned to detect such technologies, we call them “weak signals”. Usually, such technologies grow very quickly (by tens and hundreds of percent per year), but this is due to low base effect, and you have to sort through a lot of garbage to find a really worthwhile trend. Because at the stage of building up the scientific base it is not yet clear whether the technology under discussion will be the same breakthrough as AI, or whether it is another MMM. But with weak signals it is more difficult because they are very invisible against the general background, they have too low indicators.

Here’s what we found at the end of 2019:

But whether we will soon enjoy the benefits of precise medicine or elastocaloric effect – the question remains open.

More analytics on the results of Monitoring Digital Trends can be obtained from our annual reports available here. here

About plans

We will continue to quantify trends on an annual basis. In addition, this process has already been implemented in the strategic planning of Rostelecom, based on monitoring trends, we receive a list of technologies that the company plans to develop in the near future.

Every year we try to increase the amount of analyzed information, add new characteristics that affect the weights in the source, and expand the list of sources (for example, we are thinking to include the capitalization of companies, the number of article views, etc.). We are also working to improve the quality of our models and algorithms in order to minimize classification and clustering errors.

Talking about the project at various conferences, we realized that it is in demand among our partners and colleagues. Therefore, they made a product out of it for external use, which is called TeqViser… So, if such analytical tools are of interest for your tasks, I am sure that together we will be able to do something cool that no one has done before.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *