Karol Horosin says this new technology should be approached with optimism. He shares his thoughts on its challenges and talks about his experience in the software industry.

Various initiatives

The number of projects that Karol Horosin is involved in is impressive. By day, he works at Netacea, where he is involved in securing the world’s largest companies, using artificial intelligence and cybersecurity. In his own words, at the company he manages key initiatives and several teams, but also actively programs and develops new solutions. He joined the company after a previous involvement in biotechnology. At the time, he worked for a company whose mission focused on finding a cure for cancer through personalized medicine and DNA analysis. When he realized that without a Ph.D. in the field, he had limited impact on research development, he decided to take a step back and tackle strictly software projects. 

A few years earlier, Sentimatic.io, a startup that Karol co-founded, had also emerged.

“It was one of the fruits of a brainstorming session we had with a group of friends,” he says.

They then joined their first hackathon. They liked the first one so much that they quickly decided to participate in the next one. Karol notes that it was through these events that he learned to talk about projects in a way that caught people’s attention. He and his friends began to wonder what makes the idea of the prototypes they create reach people.

“The project has to be useful, tangible and have the potential for growth,” Karol concludes. – “Going to these hackathons a few ideas came up, we tried to work on a couple of them, but it was Sentimatic.io that survived.”

At first, the project focused on live media audio monitoring. These included live news broadcasts and podcasts. Karol notes that four years ago, when the project began, there was not yet the transcription for keyword analysis that there is now.

“We did this project, during the presidential election between Duda and Trzaskowski. It was very interesting. We were able to use for marketing the fact that TVN talks more about Duda than Trzaskowski, despite his political affiliation. And TVP is much more likely to say Rafał Trzaskowski than Mr. President Duda.”

Despite the interesting data collected, the project failed to sell. However, while talking to various companies, they came across an insurance company with a large call center. This company appreciated Sentimatic.io’s audio analysis, which was able to analyze emotions from recordings. This gave it access to information about calls. The company wanted to better help its customers, precisely through the better knowledge it would gain from audio analysis.

“I was skeptical at first. I thought: do you want to listen and analyze what users are saying in a call center?”, Karol admits.

The company explained that this was not the case. There are crisis situations they wanted to prevent with such technology. They gave the example of medical insurance, which a very large number of people dropped out of. People were calling the call center, saying they couldn’t get appointments, that there were no time slots. They ended up terminating the contract, Karol says. It was situations like this that the company wanted to prevent.

“So Sentimatic.io is a tool for monitoring customer satisfaction and interest in call centers, with the aim of preventing contract termination, increasing transparency and customer service quality,” Karol concludes.

In addition to the startup and his professional work, Karol also speaks at conferences and advises companies that want to implement their projects in the area of artificial intelligence.

“Thanks to the fact that I have several startups behind me that have more or less worked out, I have gone through several types of projects. It happens that someone seeing my work contacts me asking for help with a project plan,” Karol says.

Sharing knowledge

There is also a blog where Karol shares content related to his daily work.

“If I learn something and see that it works I always try to find time to describe it. So that others can find it. A lot of people get back to me with additional questions,” he says.

The blog came about, in a way, thanks to a mentor Karol came across early on in his career as a programmer. At the biotech company where he worked, he was asked to give a presentation. It concerned the creation of a user interface and was to be shown to the entire company. Karol was stressed at first, but quickly saw that sharing knowledge was something for him. 

“Then I spoke at various conferences and I wanted the content to be searchable on the Internet. That’s when I started a blog,” he explains.

He laughs that an added benefit is that while writing, his knowledge becomes structured.

Sometimes he contacts people himself, saying: “hey, I know something about this subject. Let’s talk.”

Artificial intelligence – optimism is advisable here

In an interview with Coopernicus, Karol also shares his observations on artificial intelligence (AI). He says it makes the work of programmers much easier – text and language models are great for generating code.

He sees that some senior programmers who know in which direction they want to develop their projects very often use this new technology. Karol and another programmer tried to determine the effectiveness of artificial intelligence in helping with programming. The programmer he spoke with said that as much as 90 percent of the code he writes is generated precisely by AI.

“He stresses that he spends a lot of time giving precise instructions. But it is nevertheless a huge amount of code generated by AI,” Karol says of his conversation with the programmer.

He says that artificial intelligence has the potential to eliminate a significant amount of repetitive tasks that humans do on a daily basis. It will give space to focus on other things – “on what to do, not how to do it.”

Karol has few concerns about this new technology.

“We will all be able to work more conceptually, rather than agonizing over details. That’s my hope,” he says. He adds that artificial intelligence is not yet very developed. At this point, it mainly generates text by predicting what words are likely to appear next. When you check the details, you often find that the models are not yet infallible. 

“But if you have small amounts of intelligence that can be replicated millions of times, if only by scale, these effects can be good,” he argues.

He adds that he does not share the fears often spread in the media. He also notes that young people today are not in the situation their parents prepared them for. Karol says that it is no longer the case that all one has to do is graduate from college and find a good, stable job. 

“Rather, there are no longer places where you work for 20 years at a time. What we’ve learned is that you have to adapt. For me, because I can adapt, artificial intelligence sounds interesting. Because it might be able to do a lot of things for me,” he says optimistically.

He stresses that to date, breakthroughs related to artificial intelligence also indicate that this approach is the right one. For example, the open source cluster in programming offers a lot of opportunities for development and help. He himself recently posted on such a forum, asking who would like to help him with a project to train a new language model. As many as 10 people applied, from all over the world.

“These were experienced people who were willing to give their private time to work on something new that can later be published,” Karol enthuses. – “Then there’s Facebook, which currently publishes the biggest amount of such open stuff, for example different models.”

Social exclusion can go further

One thing that puzzles Karol is the social exclusion associated with artificial intelligence. He tells Coopernicus that it’s a bit like smartphones and elderly people. Some services are only available through the phone. Although he laughs, he recently saw pensioners at a popular store using coupons on their phones.

“It was great. If they didn’t use them, it was like they were paying a bit of a tax for the fact that no one taught them to use the technology. It could be the same with artificial intelligence,” he says.

He adds that it could also be the same in the labor market. If people do not know how to use artificial intelligence, they may earn less and work more, from those who handle this new technology efficiently.

“This can also be applied to companies and corporations. If some companies have diametrically superior technology, it could further widen the gap in society,” Karol notes.

Despite this, he remains optimistic:

“There is a very long way to go before what everyone in the media is scaring us about. There is nothing to be afraid of at the moment. Artificial intelligence is useful now, it is not an existential threat to humanity.”

Artificial intelligence – challenges and how to approach them

What would he want to share with people about this new technology?

“If the beginning of a post on LinkedIn says exciting news then most likely someone generated it with an AI. A lot of my friends have such posts and I’m a bit disappointed,” – Karol laughs.

However, there is one important thing about developing and using artificial intelligence –inputting differentiated data that is free of human bias.

In an interview with Coopernicus, Karol reflects on the objectivity of artificial intelligence. 

He says that models will learn patterns from the data they are given. He gives an example:

“If we are in a city where there happens to be a gang in a particular ethnic group then of course more police reports will be about that group. But if we transfer this data blindly to another city where there is no such gang, the same data will be biased. It will be a ground where we judge someone unfairly.”

Karol says that large companies that develop language models put a great deal of effort into eliminating similar biases from the data. He says that when preparing artificial intelligence models, the saying applies: “garbage in, garbage out.”

Karol also observed a similar problem in biotechnology. He tells Coopernicus that genetic data analysis didn’t work as well on some ethnic groups as it did on others.

“That’s because simply most of the people who could afford and had access to DNA testing technology were from Europe and the United States, where most people are Caucasian, white or related ethnic groups,” Karol says.

The implication was that drugs that were created and proposed on this basis were less effective on other ethnic groups.

To counter this type of situation, projects are being developed that search for more diverse data.

“Also you have to look at the fact that the source of some of the data is the Internet. Where people have a sense of anonymity and communicate in ways they normally wouldn’t,” Karol adds, returning to the topic of artificial intelligence.

He laughs that there are a lot of data analysts who complain that it takes them a lot of time to process and prepare data. However, this is one of the most important things in the AI development process.

“We will also have companies that will cut data from the right and left of the political spectrum. But each company will have its own definition of what is extreme. So we will have models that are more conservative and more liberal,” Karol notes.

Despite all these challenges, Karol argues that we should be understanding and open to the future in working with artificial intelligence:

“We have to trust the fact that as many people live in the world, as many points of view we have. Analogously we will have artificial intelligences that will be directed in different directions and this will balance each other out.”

Artificial intelligence and business in Poland

The idea for the startup Sentimatic.io, was also born thanks to Karol’s experience with foreign language models. He noticed that the proficiency of language models in Polish is much worse than in English, for example. He tells Coopernicus that at the beginning of founding the startup, the most important thing was to create a huge data set in Polish and for Polish audio. Later, based on this data, Karol and the rest of Sentimatic.io created their language model.

Karol encourages that this is where the future lies. In his opinion, languages other than English are a niche for developing the field, and thus for doing business.

He talks about how many companies and banks don’t want to send their data to a foreign cloud. They want a good Polish language-based operation, which is not always available.

“I wanted to bring this solution to the Polish market together with the team,” he says.

He adds that for a long time Poland was an outsourcing destination for IT services. This was due to the lower cost of labor than in the United States or some countries in Western Europe.

“What distinguished us from even cheaper places like India was that people in Poland have a very large range of knowledge and skills. In addition, higher education is at a good level.”

He himself heard similar voices from people buying services from the companies he worked for. “The result is that Poles working for these companies learn. Then they have the knowledge needed to set up their own companies in the home market,” Karol notes the correlation.

He admits that he hopes this will continue to develop:

“Many people assume that their place is in our country. They have no need to go somewhere, preferring to create locally.”

He says he gets irritated when people complain about Poland. It’s easy to forget that we do a lot of good when we get lost in the negative things shown in the media or political bickering, he claims.

“The mentality is changing, people are less suspicious of entrepreneurs, there is no longer a problem with promoting success. We are not stuck 20 years ago. You can see more and more acceptance and less unhealthy jealousy,” he argues. 

Most important – real influence and cooperation

Karol admits that something that is very important in his work is having a real impact.

“I like to do things that make a difference. I know that may sound pompous, but there’s a reason why I don’t work in big companies. In startups, it’s much easier for me to realistically create products and influence what the workplace does,” he says. 

The second aspect is cooperation and having fun. This flows from Karol’s experience, who joined the AGH Space Systems research club while still in college, at the AGH University of Science and Technology in Krakow.

“The biggest strength was that it was a team that worked without pay, and we were able to implement good projects,” Karol says.

Without funds, they had to find access to materials and resources. The friendships made during the study circle also gave a lot. 

“I saw that there are people who have a lot of unused energy and you can do really great things with them. And that’s exactly how I approached Sentimatic.io. We sat down with friends, agreed that we were making a prototype for a hackathon, and we would work together,” Karol recalls. He knew he didn’t want to do the project alone, as there was a chance it would fail.

“The cooperation during my studies influenced me to not want to do projects alone anymore. Then I would lack motivation, faith in it all… And at least, together, we will have fun.”

And what would Karol advise students who are thinking about their future careers? “Do the hardest projects in your studies, take an interest in the research of the professors and offer your help, and also apply for various scholarships. This will give you a chance to work on interesting projects, go on useful trips and gain valuable experience.”

Cover photography: Unsplash