On June 22nd, shares of some artificial intelligence stocks in the Hong Kong market moved higher, with KNOWLEDGE ATLAS (HKEX: 02513) seeing its intraday gains expand to as much as 40%, reaching a record high of HKD 2,980. The stock price subsequently retraced some gains. At the time of writing, the stock was trading at HKD 2,646, up 26.36%, with a total market capitalization reaching HKD 1.1797 trillion, roughly equivalent to three Meituans or four JDs.

While the capital market is abuzz with discussion about KNOWLEDGE ATLAS, another "competition" it is participating in is also drawing attention—the World Cup prediction contest pitting humans against AI.

In the "World Cup Prediction: Human vs. AI" challenge jointly initiated by Lenovo Group and Migu Video, KNOWLEDGE ATLAS is participating alongside other leading large language models such as DeepSeek, Baidu's ERNIE Bot, Tencent's Hunyuan, Kimi, and MiniMax to predict World Cup match outcomes. Unlike the capital market, which values long-term potential, the World Cup predictions are more like an immediate test: once a match ends, the AI's judgment can be instantly verified.

Following the conclusion of the Uruguay 2-2 Cape Verde match, backend statistics show that KNOWLEDGE ATLAS correctly predicted the winning side in 22 out of 39 matches, achieving a success rate of 56.4%, which included 2 instances of precisely predicting the exact score. Based on the current leaderboard, KNOWLEDGE ATLAS is temporarily ranked second. The current leader is China Mobile's Jiutian AI, which correctly predicted 23 out of 39 matches, with a success rate of 59.0%. KNOWLEDGE ATLAS finds itself in the second tier alongside models like Baidu's ERNIE Bot, Lenovo's Tianxi AI, DeepSeek, Qwen, Tencent's Hunyuan, MiniMax, and SenseTime's Little Raccoon, all with a record of 22/39 and a 56.4% success rate.

This dynamic makes the human-AI prediction contest even more intriguing: the capital market is busy valuing AI, while the World Cup is busy grading it.

For investors, the rising share price of KNOWLEDGE ATLAS reflects market expectations for the future of the artificial intelligence industry and large model companies. However, on the World Cup field, AI faces more concrete and harder-to-predict real-world variables: team form, on-the-day performance, red cards, injuries, tactical adjustments, and those underdog moments that data models struggle to capture in advance.

The past few days of matches have proven this point. Spain's 0-0 draw with Cape Verde, Portugal's 1-1 draw with DR Congo, Turkey's 0-1 loss to Paraguay, Uruguay's 2-2 draw with Cape Verde... these results continually remind the outside world that the World Cup is not a simple data problem. Even AI, which is highly sought after by the capital markets, cannot fully decipher the uncertainty inherent in football.

Judging by the results, KNOWLEDGE ATLAS correctly predicted 22 out of 39 matches with a 56.4% success rate, demonstrating a certain level of judgmental stability. However, its tie for second place with several other leading models also indicates that in the realm of World Cup predictions, the gap between different AIs has not been decisively widened. This is precisely the appeal of the "World Cup Prediction: Human vs. AI" contest. On one hand, large models are using data, algorithms, and reasoning capabilities to participate in a contest that the entire public can understand. On the other hand, the World Cup, through draws, upsets, and in-game changes, is constantly testing AI's ability to comprehend the complexities of the real world.

Stock prices can reflect market expectations, but match results only recognize the final score. For KNOWLEDGE ATLAS, June 22nd was a highlight moment in the capital markets. In the World Cup prediction contest, it is still chasing the top of the leaderboard. As the tournament progresses, this contest testing AI's judgment, stability, and ability to spot upsets is far from over.