MakinaRocks CEO Yoon Seong-ho: “AI Transformation Depends on the Front Line for Success”
IT DAILY ·
✦ AI Summary
MakinaRocks CEO Yoon Seong-ho said in his keynote at "Attention 2026" that global frontier AI models prove their performance through benchmarks, but there is a gap between high scores and AI productivity.
He said the cause is a lack of on-site reflection, and that MakinaRocks focuses on collecting data through on-site deployment and developing AI for business outcomes.
MakinaRocks has deployed FDE teams to more than 80 sites in 32 cities across 5 countries, and it is pushing ahead with a fully autonomous manufacturing plant with Enertork, with the plan set for 2028.
At the keynote address for the "Attention 2026" conference held on the 3rd at the Westin Seoul Parnas in Gangnam-gu, Seoul, MakinaRocks CEO Yoon Seong-ho said global frontier AI models are proving their performance through benchmarks. He added, however, that there is a gap between high benchmark scores and AI productivity, and pointed to a lack of on-site reflection as the reason productivity gains have failed to materialize. The photo shows Yoon delivering the keynote speech, taken by reporter Kim Ho-jun.
Against this backdrop, Yoon said MakinaRocks collects data through on-site deployment and develops AI that can be used for business outcomes. He used the remarks to explain the company's strengths.
Meanwhile, AI technology is advancing rapidly. It took a little more than 4 years for OpenAI to go from GPT-3.5, which was applied to ChatGPT, to GPT-5.6, which can autonomously carry out even vulnerability intrusions.
Yoon, however, said the productivity gains felt on the ground have been minimal. According to a report released in May by the National Bureau of Economic Research (NBER), 90% of companies said they had not experienced measurable productivity gains over the past 3 years despite using AI.
After identifying problems in the field, Yoon concluded that even frontier models have limits in understanding on-site conditions and on-site generated data. He also said general-purpose AI falls short of what is needed for use in industrial settings where even small errors are not tolerated.
In response, MakinaRocks has focused on reflecting on-site data and work processes in AI. The company deployed an organization of front-line engineers (FDE) to more than 80 industrial sites in 32 cities across 5 countries, and those engineers logged about 70,000 hours accompanying sites this year. To date, more than 6,000 on-site implementation AI models have been created.
MakinaRocks applied an AI agent to review design document changes and cut task time by more than 1 hour. It also applied AI to more than 1,400 robots operating at 6 global automotive plants, and in the defense sector it participated in the first UFS in which AI was piloted.
The solution that represents MakinaRocks' philosophy is "Runway." Runway is a platform that supports the entire process of developing and deploying AI for enterprise sites, and it is equipped with functions for integrated management of distributed environments across data centers, factory servers, and edge devices. It also supports a range of environments, including closed networks.
The industrial application goal toward which these technologies are headed is the fully autonomous manufacturing plant on Mars that Yoon outlined. Yoon said the company's vision is to develop AI that works in the harshest and most unpredictable environments, adding that he hopes to build the world's first fully autonomous manufacturing plant on Mars.
As its first practical project for this goal, MakinaRocks is pushing ahead with a fully autonomous manufacturing plant with Enertork, an electric actuator company. At Enertork's plant in Yeoju, the company identified more than 80 tasks across both office and production operations, and it set a goal of achieving full autonomy within 3 years. Yoon said the Enertork project would serve as a starting point for realizing that goal on Earth, and the plan for the fully autonomous plant is set for 2028.
Source: IT DAILY · Kim Ho-jun
Original: https://www.itdaily.kr/news/articleView.html?idxno=241397
References
This article was produced with the help of an automated content generation algorithm.
Source: IT DAILY
View originalThis article was summarized and organized by BizCrush based on the original article from IT DAILY. For exact quotations and full details, please refer to the original article.