Stuart Russell, professor of computer science at UC Berkeley, generally known as the ‘top artificial intelligence (AI) scholar,’ attended the domestic event and emphasized ‘secure AI.’ He specifically stated that there’s a need for an 'AI Kill Switch' that may force an AI model to be shut down if it causes problems.

Professor Stuart Russell attended the sixth International Artificial Intelligence (AI) Conference 'AI Seoul 2024' held online on the Seoul City Hall Multipurpose Hall on the first and delivered this in a keynote speech.

He mentioned the potential for artificial general intelligence (AGI) and said that 'secure AI' must be created from the strategy planning stage. The logic is that if the goal is ready incorrectly, AI will inevitably produce incorrect results.

“AI should remain only a tool to attain human goals,” he argued, and said that the greater the capabilities of AI with incorrectly set goals, the greater the danger of losing human control.

Particularly, he emphasized, “If AI causes problems comparable to deep fakes, hacking, biochemical weapons production, and defamation, an AI kill switch is required to forcibly turn it off.”

As well as, AI developers are answerable for providing 'safety certification' to regulatory agencies, and international consultation is required for secure AI. “If 2023 is the yr when people wake as much as the specter of AI, then 2024 is the time to act,” he said.

In March of last yr, Professor Russell, together with Elon Musk and others, called for a ‘six-month delay in AI development.’ We now have previously emphasized ethical AI development.

On the event, Seoul National University's AI Research Institute Director Byeong-Tak Jang also appeared as a keynote speaker and attracted attention by introducing the concept of 'Embodied AI'.

Director Jang explained that because the educational speed of AI increases in proportion to computing power and the quantity of knowledge, AI performance may even improve if more data could be collected than now.

Subsequently, he explained that there are limits to accuracy because the big language model (LLM) learns only from limited text, and that AI, which has a body that may collect data comparable to vision and touch, is another.

Just as humans expand their experience through the five senses, AI may develop at a rapid pace by accepting diverse sensory data.

In a special talk following the keynote speech, Professor Russell and Director Byeong-tak Jang discussed ‘what is required in an era where AI and humans coexist.’

Professor Russell said, “We’d like a way to higher control AI,” emphasizing the necessity for international norms for secure AI, research on machine-human interaction, and enforcement agencies for regulation.

Also, when asked if there was anything he would really like to say to the residents of Seoul, he answered, “We’d like to be certain that the goals of the AI ​​model and the goals of the user match.”

Director Jang said, “Because AI has no will, intention, or value standards, necessary decisions are still left to humans,” adding, “There are concerns that AI will replace human jobs, but we must be mindful that the constraints of machines are also clear.” said.

Meanwhile, before the beginning of the event, there was a video congratulatory speech by Seoul Mayor Oh Se-hoon and Yoshua Bengio, founding father of Millar AI Research Institute.

Seoul Mayor Oh Se-hoon said, “We are going to make Seoul the middle of the AI ​​industry ecosystem through collaboration between Seoul AI Hub and Mila AI Research Institute,” and added, “We is not going to spare any investment in talent training and startup R&D.”

Founder Yosua Bengio said, “We sit up for discovering startups and collaborating on research through joint initiatives with Seoul City,” and added, “Seoul has great opportunities and potential to grow to be an AI hub.”

Seoul AI Hub is a specialized support organization within the AI ​​field supported by the Seoul Metropolitan Government and is run by Seoul National University and the Electronics and Telecommunications Research Institute (ETRI) to nurture talent within the AI ​​field and support the expansion of specialised corporations. Last December, we signed a partnership with Mila AI Research Institute and started providing support for domestic startups.

After the keynote speech, a world AI forum in the general public sector and a presentation of public innovation convergence cases using AI in domestic and foreign industries were held. Within the IR session, exchanges between industry and academia continued, including pitching by AI startups supported by Seoul Metropolitan Government and presentations of research papers within the AI ​​field.

Reporter Park Soo-bin sbin08@aitimes.com