.jpg)
Semantic Visions processes, sorts, and evaluates news content from around the world. Can you briefly explain how it does this?
Semantic Visions can search for everything that's in the global electronic media or has been published, going back up to 10 years. Our platform thus collects and analyzes data from articles in 12 languages and detects events related to specific companies or individuals or other so-called named entities. This could be commodities or products.
The fact that we at Behind Inventions invest not just finances but primarily know-how and time is clearly demonstrated in your case. How do you feel about leading Semantic Visions? And is it a temporary role?
Leading and transforming a small company is quite different from life in a corporation. (laughs) But fortunately, I knew the company and its product quite well, as we had been collaborating with them for several years. The biggest challenge is the transformation itself. First, we had to start addressing a significant technological debt, while simultaneously serving existing and new customers. On top of all that, we needed to continue innovating, as we operate in a very competitive segment where advancements in NLP (Natural Language Processing) and AI in general are rapidly evolving.
And yes, my role is temporary. My task is to kickstart the transformation, open up new customer segments, and find the right people for both the technical teams and the company's management.
What stage is the transformation of Semantic Visions for commercial customers at, and do you already have any clients in this area?
The transformation is ongoing and, given the large number of projects we are working on, it will be an endless process. In a year, we can call it only "innovation" and not compensation for technological debt. We already have several new customers thanks to changes in technology. Right now, we are most actively developing solutions for financial markets.
Do you follow trends in AI? And how do they reflect on how your system operates?
Certainly, from this perspective, Semantic Visions is an early adopter. For example, we use LLMs (Large Language Models) for classifying the role of a named entity in an event. So, when we detect a lawsuit, LLMs help us specify whether a particular company is the plaintiff or the defendant. But we also use other models in the AI field, from expert systems to machine learning and deep learning.
What's your view on government regulation of AI?
Regulation is always good and bad, depending on the viewpoint. On the one hand, it can slow down the development of this science; on the other hand, it's necessary to ensure that AI doesn't make decisions that could negatively impact people. For example, everyone likes to see AI helping to detect thieves on cameras, but at the same time, we don't like being watched. Or we like to see AI detecting pathological changes in medical imaging data more efficiently or quickly, but we wouldn't want AI to decide on the treatment itself.
It's a very complex problem that requires its investigation and a broad consensus of both professional and lay public. It takes a lot of time, which is then in great disproportion to the speed at which this field is developing.
One of the services is detecting disinformation. Are there topics you monitor long-term, or does it change according to client assignments?
Semantic Visions' service is more about detecting specific information and events. We don't interpret whether it's disinformation or not. That's up to the clients. We can accurately detect all mentions of a certain topic in different languages and provide sources. Whether it's disinformation must be assessed by organizations that have the expertise for it. Detection of topics naturally evolves because the methods and messages used in propaganda or general information operations are constantly developing.
ESG is a hot topic for companies lately; can you help in this area too?
Certainly, we currently detect up to a hundred different topics that fall under ESG – both negative and positive. We help clients, for example, monitor their supply chain for negative ESG events.
What future do you wish for Semantic Visions?
I would like us to complete the transformation to the point where we can offer processed data in a self-service mode to a much wider range of customers, and to successfully complete expansion into the UK and USA.
The interview was published in the Business Leaders magazine.
Expanding with Czech Know-How: Behind Inventions Helps Its Companies Succeed on the Global Market
