[Reporter’s Notebook] There Is No ‘Romance’ in AI Survival Games

"I’m glad I got another shot."
That was how chef Choi Kang-rok summed up his feelings after winning the recent Netflix variety show "Culinary Class Wars Season 2." The show, one of this year’s biggest hits, is framed as an intense survival competition. Twenty already-famous "silver-spoon" chefs face off against 80 "underdog" chefs from outside the limelight. The highlight of this season was not so much an underdog revolt as the passion of the silver-spoon chefs, who took off their badges of rank and poured everything into each dish, even in a situation where simply not losing was the best they could hope for. Viewers were thrilled by the winner’s humility as he downplayed himself as just an "ordinary chef" and credited the entire restaurant industry. The format was a survival contest, but the outcome was a "romantic ending" in which everyone felt like a winner.
Reality, however, is far less beautiful than a TV show. The fallout from the first-round evaluation of the Government of South Korea’s Independent AI Foundation Model Project (K-LLM), in which several teams were eliminated, has been intense. Winners and losers were clearly divided, yet the stigma attached to those who failed is drawing more attention than the achievements of those who passed. The eliminations have fueled doubts about technical capabilities and damaged corporate images. In this climate, even the government’s proposed "second-chance" round, meant as a kind of losers’ revival match, is being shunned. Naver, Kakao and NC AI have already said they will not participate. From the companies’ perspective, the risks of failing again far outweigh any potential benefits of reapplying. This is why some observers worry that the project may have lost its momentum altogether.
The Ministry of Science and ICT has explained that it "focused on developing the ecosystem rather than dividing winners and losers." Still, it is hard not to question whether such a provocative format—built on competition and elimination—was really the best approach for our AI ecosystem. Even among experts, there is no consensus on technical criteria such as what counts as being built "from scratch" or what qualifies as "proprietary." With such vague standards, the process ends up mass-producing losers and narrowing the field to just one or two final models. It is doubtful how much a Squid Game-style survival contest will actually contribute to national industrial competitiveness.
One encouraging sign is that AI startups such as Motif Technologies and Trillion Labs have expressed their intention to try again. They did not make the initial list of "national AI representatives," but they have quietly honed their technology and built impressive models on their own. The Government of South Korea now needs to reassess the rules of this survival game. The goal should not simply be to pick a single winner, but to design a win–win structure in which all participating companies can grow together. The hope is that, instead of saying "It was excruciating to be eliminated," companies will one day be able to say, "I’m glad we took on the challenge."
wongood@fnnews.com Reporter