플레이어로 건너뛰기본문으로 건너뛰기
The team with the best benchmark score was the one eliminated from Korea's sovereign AI program.
Upstage, SK Telecom and LG AI Research cleared the second-stage evaluation of Korea's independent foundation model project, while Motif Technologies finished last despite the strongest intelligence index score, losing on real-world usability. The user evaluation category produced the widest spread of any criterion, and the ministry is now considering a redesign that moves away from selecting only a few winners. The episode also covers Japan's JPYC stablecoin payments reaching over 100 merchant inquiries on zero fees, and a global first-half IPO market that tripled to 205.1 billion dollars.

Sources:

Upstage, SKT and LG in Three-Way Race as Real-World AI Decides — Seoul Economic Daily, August 18, 2026
Paying for Drinks With Yen Stablecoin, Zero Fees Draw 100 Inquiries — Seoul Economic Daily, August 18, 2026
Global IPO Proceeds Tripled, but 42% Went to SpaceX — Seoul Economic Daily, August 18, 2026

About AI PRISM:
AI PRISM is Seoul Economic Daily's WAN-IFRA award-winning newsroom AI series, delivering Korean economic news adapted for global audiences. Episodes are produced with AI assistance and reviewed by a human editor.

Tags:
#Startups #AIFoundationModels #Upstage #Stablecoin #JPYC #IPO #KOSPI #AIPRISM #SeoulEconomicDaily #WANIFRA

카테고리

🗞
뉴스
트랜스크립트
00:02You're listening to AI Prism from Seoul Economic Daily.
00:32It's Wednesday, August 19th.
00:34Let's look at what this means for startups and entrepreneurs in Korea.
00:39Korea's Sovereign AI Foundation Model Program ran its second-stage evaluation this week.
00:44The ministry scored teams on benchmarks, expert review, and actual user experience.
00:50The category that decided the outcome was not raw capability.
00:54Here's what the numbers show.
00:55Upstage SK Telecom and LG AI Research passed the second stage,
01:01and Motif Technologies finished last and was eliminated.
01:04The user evaluation, worth 25 points, averaged 17.6,
01:09and produced a 5-point spread between 1st and 4th place, the widest of any category.
01:15The benchmark category, worth 40 points, averaged 22.5 with a maximum gap of 4 points,
01:22and expert review, worth 35 points, averaged 28.8 with a gap of 2.4.
01:27Testing drew on 49 professional users, including AI startup executives and 185 general citizens.
01:36Upstage's SolarOpen 2 reached a 1 million token context window,
01:40and LG's KXA1 2.0 runs 750 billion parameters, the largest in Korea.
01:49In Tokyo, the JPYC stablecoin platform Misipay charges no payment or setup fee
01:55against card fees of about 25,000 yen on 1 million yen in monthly volume,
02:00with JPYC circulation near 2.73 billion yen.
02:04So, what does this mean for founders choosing where to compete?
02:09Earlier, we said both stories point at the same thing.
02:12Here's what that actually means for you.
02:14In Korea's largest AI funding program, deployed usage outweighed benchmark scores,
02:20and the ministry is now redesigning the program so it stops picking only a small number of winners.
02:26Build the distribution case into the pitch because that is the axis the government just scored.
02:31Watch the third-stage evaluation, which narrows three teams to two as early as the end of this year.
02:37Watch the Ministry of Science and ICT announcement on the redesigned support structure,
02:43due once the budget is confirmed.
02:45Watch the second-half IPO window after global first-half proceeds triple to $205.1 billion,
02:52with 42% going to SpaceX alone.
02:55That's today's AI Prism founders.
02:58This episode was produced with AI assistants based on Seoul Economic Daily reporting
03:03and reviewed by a human editor.
03:06AI Prism is a Juan Ifra award-winning series.
03:09We'll be back tomorrow.
03:10You've been listening to AI Prism from Seoul Economic Daily.
03:14I'm the USDIY.
03:14Bye!
03:15Bye!
03:15Bye!
03:15Bye!
03:16Bye!
03:17Bye!
03:17Bye!
03:17다음 영상에서 만나요.
댓글

추천