Dr. hab. Przegalińska: over the past 2.5 years, there has been a revolution in AI availability

Since the premiere of Chata GPT 2.5 years ago, a revolution has taken place: artificial intelligence has become maximally accessible and there are no barriers related to user competences - Dr. hab. Aleksandra Przegalińska from the Department of Management in the Network Society at the Kozminski University (ALK) told PAP.

"Many people may be surprised, but artificial intelligence has been developing for 80 years," noted Aleksandra Przegalińska, professor at ALK. "It is a field in which new algorithms are constantly being created, which can learn from the database and improve what they do," she added.

She reminded us that there are many subfields within artificial intelligence. "There are algorithms that power robotics; there are machine vision algorithms that allow us to recognize objects around us. There are so-called predictive algorithms that are responsible for a good weather forecast or perform analytics on potentially suspicious bank transactions. Recently, a very 'hot' field of artificial intelligence has been the so-called generative AI - these are also algorithms that operate probabilistically, which are able to generate, produce something. They can produce, among other things, text, visualization, image, musical composition, video. They are largely similar to these other predictive algorithms, but they also have the property of producing something new based on data," said Prof. Przegalińska.

The expert emphasized that this is a huge change: "Not only because the algorithms perform their tasks quite well - although they have their limitations; but because in principle this is the first wave of algorithms with which you can calmly talk; which do not need to be programmed; where these technical skills (on the part of the user - PAP) may be at a much lower level, while the skills that are natural to us - conversations - can be used to their full potential. Everyone can talk to the GPT chat or another model and get something from it."

The ALK professor noted that in the two and a half years since the premiere of the GPT hut, not only have many more models appeared, but there has also been a revolution in accessibility. "This technology has become as accessible as possible; there are no barriers related to competences," she said.

As a result, this technology is used on a very large scale.

It is used by professionals, for example in creating programming code. "When it comes to the work of programmers, this change is really very big. There are professions in the area of ​​programming that are disappearing before our eyes," the expert pointed out. "On a slightly smaller scale, because they are not as perfect - these models are used in other areas, for example in management, in creation, in creativity," she added.

Przegalińska pointed out that the great challenge related to AI concerns the education sector. "It is known that students use these models when learning. And this is OK; it is a kind of tutor for them who helps them prepare. But on the other hand (students) simply assign this artificial intelligence to do things instead of them. And this means that they ultimately lose out because they will not learn something, because they will transfer some task to the technology to perform. This is a very big challenge: how to deal with it? How to verify whether someone really knows something and check their work with such democratization of these tools?" - wondered the expert.

In recent months, there has been increasing talk about the problem of generating articles using AI and then publishing them in reputable scientific journals without any indication that artificial intelligence was involved in their creation.

As Przegalińska emphasized, such actions do not serve anyone in the scientific community. "If we allow ourselves to produce knowledge that is not knowledge, but is a hallucination - it will hit us all," the researcher concluded. She suggested that scientists must verify in the coming years "where this technology really provides value, and where it is only a springboard to the next career rung, to 'score' publications." "Because this is a very weak strategy that does not build good science," she concluded.

She noted that AI should be used as a very valuable tool, e.g. in chemistry, physics or social research: "there is definitely a place for it there."

"I think we are slowly seeing a correction of sorts. Understanding that anyone can generate any article these days, as a scientific community we are responding to this with increased verification," she added.

Aleksandra Przegalińska took part in the Perspektywy Women in Tech Summit (June 4-5, Warsaw). The Polish Press Agency took the event under its media patronage. (PAP)

Science in Poland

an/ arw/ bar/ pl/