MINSK, 23 December (BelTA) – When working with artificial intelligence, it is essential to provide the right context and to formulate prompts properly, Belarusian Communications and Informatization Minister Kirill Zalessky said at the 12th Forum for Young Journalists Beyond Clickbait, BelTA has learned.
“Artificial intelligence is neither good nor bad. It’s a phenomenon of our era that we must learn to manage. By default, AI isn’t built to admit ignorance. Instead, it will generate the statistically most probable responses and steer you in the wrong direction. Proper context is key. The higher the quality of our prompt, the higher the quality of its answer,” Kirill Zalessky said.
Kirill Zalessky added that AI systems are based on digital systems containing mathematical elements. They do not formulate an answer based on meaning. Instead, they mathematically select the next most suitable word depending on all previous ones. Every prompt is split into tokens. And that’s not necessarily a word. A token can be just a part of a word, a letter, or a syllable. Each token is selected based on statistical probability compiled from analyzing all possible information on the internet that the AI could access during its training. This is precisely why artificial intelligence sometimes generates “hallucinations”: convincing information that is, in fact, incorrect.
“Artificial intelligence is neither good nor bad. It’s a phenomenon of our era that we must learn to manage. By default, AI isn’t built to admit ignorance. Instead, it will generate the statistically most probable responses and steer you in the wrong direction. Proper context is key. The higher the quality of our prompt, the higher the quality of its answer,” Kirill Zalessky said.
According to him, young people often trust modern technologies without fact-checking and aren’t even aware of the pitfalls embedded within. And this is a fundamentally important point for a journalist, who must bear responsibility for what they write and say publicly, as a journalist influences people’s minds.
Kirill Zalessky added that AI systems are based on digital systems containing mathematical elements. They do not formulate an answer based on meaning. Instead, they mathematically select the next most suitable word depending on all previous ones. Every prompt is split into tokens. And that’s not necessarily a word. A token can be just a part of a word, a letter, or a syllable. Each token is selected based on statistical probability compiled from analyzing all possible information on the internet that the AI could access during its training. This is precisely why artificial intelligence sometimes generates “hallucinations”: convincing information that is, in fact, incorrect.
“We often hear questions whether artificial intelligence will replace humans. From my generation’s perspective, no, because the final decision must be made by a human. I don’t know the approach of the future generation. Perhaps, given that in some areas artificial intelligence can be more accurate than a human, the new generation will reach an internal societal consensus on this issue,” Kirill Zalessky emphasized.
Photos by Tatyana Matusevich/BelTA
Follow us on:
