While playing LiverLive with friends over the weekend, I discovered a blind spot in the AI. They can't distinguish between 4/3 and 4/4 time. They're blindly guessing that there wasn't relevant data during pre-training, or at least the data for Chinese songs is missing. So they're just making random guesses and getting almost all of their predictions wrong (both GPT and Grok did the same; they all gave 4/4). GPT Chinese Songs Later, out of curiosity, I tested it and found that it was indeed the lack of Chinese data. The English song recommendations were all correct, indicating that they can still distinguish them. GPT English Songs Then I tested Deepseek to see if the Chinese large-scale model could incorporate this data during pre-training. The results were indeed better than GPT and Grok, but only half as good, as half of the results were 3/4 (or 6/8) and the other half were still 4/4...but it's still better than the two foreign models that got all of their predictions wrong. Deepseek Chinese Songs (I'm too lazy to test English + (Image limit: 4) Finally, I recommend the LiberLive Entertainment Electric Guitar. It's fun and affordable, perfect for those with little musical experience who want to learn to play and sing. Getting started in 10 minutes is comparable to the initial 1-year launch on the mainnet and the 5-minute one-click link launch we're experiencing now. :)