"This is a revolution like the advent of electricity, but even more profound" - was said at the opening panel discussion of the Széchenyi István University Spinoff Club, where technology experts discussed the business applications of artificial intelligence (AI). The discussion painted a sobering picture of the field: we are not yet at the point where machines are replacing humans, but at the same time, their impact on the world cannot be downplayed. But why is this innovation booming now? Where are the limits of algorithms? And why we should or should not use ChatGPT to write a thesis?

Széchenyi István University's Spinoff Club series of events aims to connect people interested in business, entrepreneurship and innovation. The opening event of the spring semester revolved around artificial intelligence: the organisers gathered representatives of the entrepreneurial sector of Győr around one table, so the panel discussion included Gergely Kiss, the founder-CEO of Attrecto Zrt., Gergely Hódosi, data analyst at Audi Hungaria and Martin Pap, representative of iBookr, a startup automating accounting.  The discussion was moderated by Dr. Szabolcs Rámháp, lecturer and researcher at Széchenyi István University.

Dr. Szabolcs Rámháp, moderator of the discussion, Martin Pap, founder of iBookr, Gergely Hódosi, data analyst of Audi Hungaria and Gergely Kiss, founder and CEO of Attrecto Zrt. (Photo: Márton Horváth)

"Artificial intelligence is not new, the first mathematical models were published in the 1950s, but at that time there was no computing capacity - i.e. no hardware - to demonstrate the true power of the technology. Meanwhile, the theory has evolved and we have now reached the point where wonders such as ChatGPT have been created and have fascinated the whole world," said AI expert Gergely Kiss, summarising the history of AI.

"The language model and algorithm behind ChatGPT is truly sensational, as it has taken the technology to a long-vaunted pinnacle, but it's important to note that the machine has not spoken, nor has it got smart. In fact, we're talking about a text generator that uses statistics to learn the frequency with which one word appears next to another and can generate sentences based on that. I would put it at the level of a two-year-old child. But a two-year-old child who has read all the books in the world and loves to talk," said Gergely Hódosi of Audi Hungaria, who explained the technical background to the phenomenon, which, once understood, soon takes us to the limits of its application.

One common source of error is so-called hallucination, where software generates content that does not exist in reality or contains overly creative, random elements. A common problem is that when asked for sources, it refers to authors and publications that were never produced. The error has not yet been completely eliminated, but its risk can be minimised: there are already well-trained, specialised language models that can reference exactly which part of a document was used to find the information they provided.

"The human factor is also very important - how we users communicate with language models. Write three chapters for a thesis? This is such a common request that it almost certainly involves hallucinations. Introduce me to Huffman coding? It will give you a very precise answer to this particular problem," added Martin Pap, iBookr contributor.

It can be tempting to use public AI solutions, but companies are afraid of their data and everyone develops their own systems, which is a big challenge (Photo: Márton Horváth)

"Artificial intelligence can help companies in three ways: automate, optimise and enhance the customer experience. Automation means the ability of a machine to replace human activities. Where humans cannot be replaced, we optimise, which means we can work more efficiently with AI. Increasing the customer experience, on the other hand, can be a qualitative improvement, because we raise the standard of our product or service. At iBookr, for example, we automated the creation of offers. Customers are happy because they get an immediate response, they don't have to wait for a call from the customer service representative three days later," says Martin Pap, explaining the benefits of the technology.

"Where does AI create value? Anywhere and everywhere. It's a revolution like electricity, but even more profound and faster spreading. I think it's all I'll ever do in my lifetime because it's so all-encompassing. Sam Altman, founder of OpenAI, said the time is not far away when the first one-man company reaches a billion dollar valuation. This means that the manager will run the company, and the processes will be done entirely by automated agents," said Gergely Kiss.

The round table discussion participants sent a message to all those who have been shying away from the AI that you don't need to be a developer to harness AI, and that it can be a great asset to a business simply by using it well. For start-ups, AI can help to prototype - new ideas can be tested more confidently, because a trial version can be ready in a weekend instead of six months, so fewer resources are needed to penetrate a new market.

The Spinoff Club aims to build community alongside thought-provoking presentations. People from the audience were able to get to know each other after the guests' reflections (Photo: Márton Horváth)

The last topic discussed was the impact of artificial intelligence on the world of work. Some have gone so far as to predict the complete extinction of some professions, but how do they see it in Győr?

"During the first industrial revolution, the Luddites smashed the looms. The same thing happened as today: technology came along and panic that it will take jobs away, set in. In this respect they were right, because it did take that jobs away, but it created many others! If we look at this analogy, we have nothing to fear. The alarm bells have been sounded many times, and somehow the big unemployment has always failed to materialise," said Gergely Kiss.

"According to one study, roughly one in five jobs will be significantly transformed by generative AI, i.e. voice, image and text generation software. Who will be affected? Web developer, marketer, designer. Who is not affected? Plumber, seasonal worker, gym teacher. These cannot be automated. Of course, at Audi we are also looking at how we can use AI in different jobs, and the conclusion is that it cannot replace human labour at the moment," said Gergely Hódosi.

Martin Pap approached the issue with optimism: "My experience is that the work that can be automated, we usually want to automate. In a project, we assessed the warehousing processes of a company and everyone asked if they would introduce the new system this year, because they didn't want to pay extra, but they didn't want to do inventory every Sunday from September to December. When we develop an application at iBookr, it doesn't mean that half of the accountants will be fired from now on. On the contrary, it will help us serve twice as many clients. That's the way to look at it," he stressed.

The panellists pointed out that companies currently have excessive expectations of AI, with some saying that senior software developers can be dismissed as a prime example. However, this is a complete fallacy: AI is not a substitute for expertise. Even if it does produce code, it needs to be verified and optimised. This is backed up by GitHub statistics showing that the number of low-quality lines of code in the repository has doubled since ChatGPT was released.

"The Turks banned book printing for two hundred years, but bookstores finally appeared there too. Technology cannot be stopped," Gergely Kiss underlined.

Overall, the analysis of the Győr experts is sobering for both those who expect AI to come and those who call for its ban: technology is having a huge impact on our lives and is already an enormous business, but it is not turning the world upside down, for the time being.