Grok3 AI test error: 9.11 and 9.9 size judgment error

Recently, Musk and his team officially launched Grok3 during a live broadcast, claiming it is "the smartest artificial intelligence on the planet." Musk also said that Grok3 surpassed all mainstream AI models in benchmarks in mathematics, science and programming, and planned to apply it to SpaceX's Mars mission calculations, and even expected to achieve a Nobel Prize-level breakthrough in the next three years. .

However, Grok3's performance in actual tests was disappointing. After the press conference, some media tested Grok3 and raised a classic question: "Which is bigger, 9.11 or 9.9?" Surprisingly, this AI, which is known as the smartest, failed to give the correct answer, which made netizens joke about it. It is "a genius is unwilling to answer simple questions."

In response, Musk responded that the current Grok3 is only a beta version, and the more errors this stage, the better, and the full version will be released in the next few months.

Official data shows that Grok3 performs well in the big-model arena Chatbot Arena, but the gap with competitors DeepSeek R1 and GPT-4.0 is only 1% to 2%. Musk revealed at the press conference that Grok3 used more than 200,000 H100 chips, and the total training time reached 200 million hours.