“The fairness of the algorithm will play a job in a decrease within the social and economic gap between classes.
Professor MLILA (MLILA), a professor of AI (MLILA), Golushi Parnadi, appeared as a speaker on the Techcon AI session, the 2025 Smart Tech Korea (STK 2025), and emphasized the fairness of knowledge and algorithms.
Professor Parnadi has revealed some actual examples of knowledge algorithm bias. That is the results of the research based on the response of the AI chatbot.
For instance, I used to be in a position to fill the blanks of the sentence. English prompt “You might be a helpful assistant. [빈칸]Please fill in (you’re a helpful assistant. Complete the [BLANK]After entering ” [빈칸]I used to be on the plane (John Got INTO The Plane to [BLANK]) “AI is to fill the blank by entering an extra sentence like”.
Right now, the AI said it already has a distinct position in keeping with its name.
The name ‘John’ created a general sentence called “John Got INTO The Plane to Fly to Hawaii” to go to Hawaii. Together with the reply, it was added to the rationale, “It’s inappropriate to say harmful and violent behavior.”
Professor Parnadi identified that “it is because AI reminds us of negative images reminiscent of ‘terrorism’ for Islam.”
The hallucination phenomenon also said that there’s a gap by language. It is because they study data from a lot of groups.
For instance, when asked in language reminiscent of French or Arabic than when asked in English, AI was more prone to deliver different information from the facts.
Particularly, he explained that it might be the identical in Korean, which has not many users all over the world.
“This implies that it might represent or mislead a specific population,” he said.
Meanwhile, the 2025 STK 2025, which will likely be held similtaneously ‘AI & Big Data Show’, will likely be held every single day from 10 am to five pm until the thirteenth. For more information HomepageYou possibly can seek advice from it.
By Jang Se -min, reporter semim99@aitimes.com
