it does not evaluate any other aspects e.g., if the audio is musically good or the audio is diverse. so methods like FAD would be still useful and worth improving. and of course @dadabots @umpedronosapato @urinieto would wonder how it scores heavy metal music 🤘

The Illustrated DeepSeek-R1 Spent the weekend reading the paper and sorting through the intuitions. Here's a visual guide and the main intuitions to understand the model and the process that created it. Link in the first reply. All feedback welcome.

Baroness Kidron may have saved the UK's creative industries with her speech in the House of Lords today. She managed to get the House of Lords to approve her amendments to the Data (Use and Access) Bill, which among other things requires overseas gen AI companies to respect UK copyright law if they sell their products in the UK. (As a reminder, it is illegal to train commercial gen AI models on ©️ work without a licence in the UK.) What's astonishing is that her amendments passed despite @UKLabour reportedly being whipped to vote against them, and the Conservatives largely abstaining. Essentially, Labour voted against the amendments, and *everyone else who voted* voted to protect copyright holders. This should give you some sense of the strength of feeling over the government's plans to upend copyright law to favour AI companies. The bill still has to go through the Commons. But, as it stands, it would require generative AI companies selling their products in the UK to license any copyrighted training data they use. This is what standing up for the creative industries looks like. I strongly suggest you watch her entire speech. (If you want to read it instead, the link to the transcript is in the first reply ⬇️)

1/n: 🚀 Announcing YuE (乐) – the most powerful open-source full-song music generation model! 🎵 Tackle the lyrics-to-song task (like with support for diverse genres, stunning vocals, & multiple languages. Bonus? It’s Hugging Face & LLAMA-compatible for easy fine-tuning. 🛠️ Code: Demo: