JAKARTA - The 2026 World Cup is not only a stage for players and coaches. Off the field, a number of Chinese artificial intelligence models compete to guess the results of the match.
Based on a report by China Daily, which was quoted on Tuesday, June 16, a number of large language models or large language models (LLM), such as Qwen, DeepSeek, Kimi, and MiniMax, launched a World Cup prediction feature. This tournament is an arena to test AI's ability to read data and analyze match opportunities.
The 23rd edition of the World Cup was attended by 48 teams and was held in the United States, Canada, and Mexico. The tournament opened on Thursday and runs until July 19.
Guo Tao, a member of the Chinese Association for Artificial Intelligence, said the World Cup provides a rare opportunity for AI companies to show their computing and analysis capabilities to a wide audience.
"As one of the most watched sporting events in the world, the World Cup gives AI companies a rare opportunity to show their LLM computing power and analytical capabilities to a wider audience," Guo said, as quoted by China Daily.
Some platforms create interactive campaigns. Kimi owned by Moonshot AI, for example, launched a collection of 1 trillion token rewards. Users can share rewards if they successfully guess the winner of the match and the final champion.
Tokens are the smallest unit of data processed by an AI model. Meanwhile, Alibaba Group's Qwen presents a special assistant for predicting matches and human prediction challenges against AI.
However, the World Cup also showed the limits of AI's ability. Before the Group C match between Brazil and Morocco on Sunday, a number of major LLM recommended Brazil based on historical data and statistical indicators. The result, the match ended in a 1-1 draw.
Guo said AI is indeed able to read old data and statistical models. However, sports are still difficult to predict because they are influenced by many factors in the real world that are not easily measured by fixed models.
The limitations were also highlighted by Wang Zhongyuan, President of the Beijing Academy of Artificial Intelligence, at the BAAI Conference last week.
According to Wang, LLM is increasingly able to solve problems in the digital world. However, many problems in the physical world are still difficult to reach. Therefore, the direction of AI development will shift from "predicting the next token" to "predicting the next physical state".
Although the accuracy is not perfect, technology companies are still entering sports predictions. Guo assessed that this step is inseparable from the tight competition in the AI industry.
"As competition in the LLM market becomes more intense, technology differentiation becomes more difficult. Companies are eager to find new channels to differentiate themselves from competitors," said Guo.
He said the market is no longer just looking at the size of the AI model. More importantly is whether the model can provide real services and help solve user problems.
Hu Yanping, a professor at Shanghai University of Finance and Economics, said LLM and AI agents are beginning to shift from conversational systems to systems capable of carrying out tasks. The development also leads to continuous learning and a broader understanding of the real world.
"Exploratory projects, such as World Cup match predictions, can help accelerate this evolution," Hu said.
According to Hu, future AI agents need capabilities built on perception, interaction, decision-making, and collaboration.
The English, Chinese, Japanese, Arabic, and French versions are automatically generated by the AI. So there may still be inaccuracies in translating, please always see Indonesian as our main language. (system supported by DigitalSiber.id)
