Karen Hao (far left), a contributing writer for The Atlantic, speaks during a discussion session, titled "The Rise of Super-Intelligent Agents," at the EmTech Korea conference held in Seoul on Thursday. On her left is Yoon Kim, a partner at Saehan Ventures; Tong Zhang, a professor at the University of Illinois Urbana-Champaign; and Lee Moon-tae, a fundamental research leader at LG AI Research. (Jie Ye-eun/The Korea Herald)
As humans are increasingly associating with artificial intelligence at home and in the workplace, global tech experts gathered Thursday at this year's EmTech Korea to discuss what these technologies might look like in the next decade.
Orchestrated by the MIT Technology Review, the annual conference aimed at exploring the fusion of global technology trends and industries took place at the Coex convention and exhibition center in Seoul for the first time.
In a panel discussion moderated by Karen Hao, a tech writer, the three panelists -- Tong Zhang, a computer science professor at the University of Illinois Urbana-Champaign; Lee Moon-tae, a lab leader at LG AI Research; and Yoon Kim, a partner at Saehan Ventures, talked about the rise of superintelligence agents and the changing nature of human-machine collaboration.
A superintelligence refers to a hypothetical AI or agent that can surpass human intelligence across the most economically valuable or intellectually demanding tasks. Some examples of artificial superintelligence include chatbots and AI tools such as Apple’s Siri, OpenAI’s ChatGPT and AlphaGo.
Tech professionals said that superintelligence agents will make humans’ lives much easier in general by doing things more accurately, efficiently, naturally and safely -- but there are also limits to their capacities.
Before the panel discussions, in the session dubbed, “Large Language Models and the Prospects for Generative AI," Zhang briefly introduced types of large language models and artificial general intelligence alongside his ideas on how they will likely change our lives in the next 5 to 10 years.
First, he suggested that generative AI for search would look different in the next decade, citing the example of the AI search engine Perplexity, which will give users summarized answers with attributed sources.
While artificial general intelligence refers to a type of AI that can understand, learn and apply knowledge across a broad range of tasks, much like humans can, LLMs' performance would become more important in the AGI system, according to the expert.
In his presentation, Zhang forecasted that humans will see much more reliable AI agents in the next five years, as developers require feedback and cross-checking processes to reduce false results.
“We’ll see a lot of development in the next 10 years so that (agents) can actually understand physical work to enough extent that they can work with you. You will see the real intelligent robots and physical robots roaming around,” Zhang said.
Although human-machine collaboration could make life much more convenient, people cannot see themselves fully depending on the agents in the next decade, experts said.
"AI agents require extra human effort to utilize data and feedback for continuous machine learning even after ten years since large language models do not have memory span like humans," Zhang added.
Meanwhile, a participant from the floor asked the experts what types of efforts humans and AI should make to preserve humanity, as AI agents will likely outnumber humans in the future.
"Humans really have a challenge of trying to think about what really makes humans humans. Certainly, AI is not going to help teach us how that is. I think humans need to understand themselves in that new realm of knowledge and sphere of information. I am very confident that humans will come up with very new, very exciting, unbelievable things that AI will never be able to make," Kim said.
MOST POPULAR