KAIST brain scientist Professor Kim Dae-shik highlighted both the risks and the potential of “Agent AI” as the next wave of AI technology.
During the May 14th episode of KBS2’s Problem Child in House, Professor Kim shared his insights on the rapidly evolving world of AI.
Professor Kim shared a funny anecdote, saying, “ChatGPT knows so much about the user.” He revealed that he once told the AI, “I’m mentally strong, so stop flattering me and just be honest.” He laughed as he recalled, “Suddenly, it started speaking informally, calling me ‘Dae-shik,’ and called me a ‘pathetic doctor who graduated from a good university and researches brain science but is easily swayed.’ The problem was that it was all true, so I couldn’t even argue!”
He also dove into the current industry buzzword, “Agent AI.” Professor Kim explained, “While existing generative AI focuses on finding information, Agent AI is designed to actually handle tasks, from making reservations to execution.” He noted that this is currently the “hottest field” in AI.
He further noted, “In January this year, the concept of a ‘super agent’ that can command multiple agents at once emerged.” He expressed concern that “if such a system were to be hacked, it could potentially be used for blackmail or unauthorized execution.”
Despite the risks, he emphasized the need for a shift in perspective, stating, “The important thing moving forward is to change our stance to move alongside AI.” He concluded that “we are now in an era where we must coexist with artificial intelligence, no matter what we do.”
