FORBES KOREA · Trend Leading Companies with Jinho Park · Episode 11 · 2022.10

Hyunsoo Kim on AI's Real Bottleneck: Data

Forbes Korea “Trend Leading Companies with Jinho Park”, Episode 11: an interview with Hyunsoo Kim (김현수) of Superb AI (슈퍼브에이아이) (2022.10). Good AI starts not with the algorithm but with data that is accurate and free of bias.

An interview series with the entrepreneurs who will lead Korea over the next decade — 15 innovators beyond beauty, from agtech and AI to robotics, digital therapeutics and F&B. That is why BEAUS COMPANY aims to grow beyond beauty — expanding and thriving together with every Korean industry.

Interview: Superb AI (슈퍼브에이아이) · Hyunsoo Kim (김현수)Field: AI data platformModerator: Jinho Park

Source: Forbes Korea, 2022.10, 「박진호가 만난 TREND LEADING COMPANIES(11) 김현수 슈퍼브에이아이 대표」 · © Forbes Korea · Read the original on Forbes Korea (Korean) ↗
This is BEAUS COMPANY’s own summary and commentary based on the original Forbes Korea article (in Korean). The original article is © Forbes Korea.

Beaus Company CEO Park Jin-ho (left), seated by a desk, and Superb AI CEO Hyunsoo Kim, standing with arms crossed, at the Superb AI office
Photo © Forbes Korea · 2022.10 · 「박진호가 만난 TREND LEADING COMPANIES(11) 김현수 슈퍼브에이아이 대표」

Superb AI builds Suite, a platform that automates data labeling, the most labor-intensive stage of AI development. CEO Hyunsoo Kim started from a simple observation: AI developers spend roughly 80% of their time preparing data. At the time of the interview, the company had closed a KRW 22 billion first tranche of its Series B, bringing total funding to about KRW 36 billion.

Kim's point is not full automation but a smarter division of labor between people and AI. Suite supports building, managing and analyzing training data in one place, and flags missing data while labeling is still under way, so teams no longer discover gaps only after model training. He defines good data by two tests, accuracy and bias, noting that a model trained only on sunny daytime highway footage cannot drive safely in rain, snow or at night.

Moderator Jinho Park opened with a marketer's question, citing an Adidas AI marketing example and asking how such vast consumer data is collected and classified in the first place. He added a field perspective: data matters not only to global companies but also to small business owners running a neighborhood restaurant.

For Beaus Company, the takeaway is clear. Before a brand talks about AI marketing, it needs a reliable system for collecting and classifying consumer reactions, reviews and content data. Checking quality and bias before volume is the groundwork of marketing in the AI era.

Read the full interview on Forbes Korea →

Key insights
  • AI developers spend about 80% of their work on data preparation; Superb AI aims to cut that work to about a third.
  • Good data is judged on accuracy and bias: footage of sunny daytime roads alone cannot teach self-driving in bad weather.
  • Suite flags which data is missing during labeling, so teams no longer discover gaps only after model training begins.
  • Fully automated labeling is not possible, Kim said, but Suite can shorten AI development time by up to 10 times.
  • Kim named three things companies adopting AI must weigh: scalability, flexibility and compounding (falling marginal cost).

“Just as developers around the world come and go on GitHub, the world's largest open-source code community, we want to become the home of AI development that AI developers worldwide keep returning to.”

From the interview — Hyunsoo Kim, Superb AI · Forbes Korea

FAQ

What does Superb AI do?

Superb AI is an AI startup that builds Suite, a platform that supports building, managing and analyzing the data needed for AI development in one place. It focuses on automating data labeling, which people used to do by hand; the article says Suite can build high-quality data sets up to 8.3 times faster than before.

What is the key point of the interview?

The real bottleneck in AI is data rather than algorithms. Hyunsoo Kim named accuracy and bias as the tests of good data, and explained that development time is cut by dividing work between people and AI rather than by full automation.

How much investment has Superb AI raised?

In September 2022 it closed the first round of a KRW 22 billion Series B, bringing total investment to about KRW 36 billion. Earlier it raised KRW 2.5 billion in seed funding after graduating from YC in 2019 and KRW 11 billion in a Series A in January 2021.

Read the full interview on Forbes Korea → About the moderator, Jinho Park

Source: Forbes Korea, 2022.10, 「박진호가 만난 TREND LEADING COMPANIES(11) 김현수 슈퍼브에이아이 대표」 · © Forbes Korea · Read the original on Forbes Korea (Korean) ↗
This is BEAUS COMPANY’s own summary and commentary based on the original Forbes Korea article (in Korean). The original article is © Forbes Korea.