The post discusses recent challenges aimed at testing the reasoning capabilities of large language models (LLMs), specifically criticizing the notion that the tests genuinely measure reasoning rather than memory recall. Users have noted that the problems presented seem to rely heavily on the model's ability to identify known entities rather than engage in complex reasoning processes. One user pointed out the limitations of the tests when compared to human cognitive abilities, arguing that they might not accurately reflect reasoning skills. There is also a concern about the potential biases in the training data of the models, as the effectiveness of LLMs can be impacted by what was included or omitted from their training corpus. Overall, while the intention behind the challenge is noteworthy, its execution may not truly capture the nuanced aspects of reasoning that are desired. Furthermore, discussions around the educational attainment, such as the PhD, highlight the real-world problem-solving skills acquired through experience, suggesting that rote knowledge does not equate to intelligence.