Key Highlights
- Hualien AI Music Lab launched with human-AI musical performances
- Features Fed-GPT technology with 100 billion file database
- Raises ethical questions about data sources and artist compensation
The grand opening of the Hualien AI Music Lab last weekend wasn’t just another tech launch – it was a glimpse into a future where AI and cultural heritage create something entirely new. With the motto “Inspire the world by singing your song,” this lab is reimagining how music is created, shared, and preserved.
Housed in a century-old building that was formerly the H&W Winery, the lab spans two floors and has been transformed into a multifunctional space. Beyond performances, it offers recording facilities, group practice rooms, and classrooms for discussion and teaching.
If you closed your eyes during the opening ceremony, you’d have been hard-pressed to tell which notes came from humans and which from machines. Operations director Ado, an Amis singer, performed alongside an AI voice with such precision that the distinction between human and digital became meaningless. She stated “The conventional wisdom of our ancestors is our weapon. We will embracethe future with Taiwan’s most proud new technology.“
Music director Suming also showcased this technology by singing with his AI counterpart.
Screenshot
Several prominent Taiwanese musicians attended the event, including Lim Giong, Li Jian-fu and Coco Fong, highlighting the music industry’s interest in this technological advancement.
At its core, the lab runs on Fed-GPT technology with a 100 billion file database. This technology can analyze Taiwan’s pop music videos contained in its extensive database. The lab will feature regular performances by AI-generated singers Huia and Juan, created by Taiwan AI Labs using the voices of real singers.
According to a recent Goldmedia report, 95% of music creators surveyed want AI providers to disclose when they use copyrighted works as training data.
- The Goldmedia report also notes that 90% of creators believe they should benefit financially when their works are used as input for AI training.
- Similarly, 90% think permission should be obtained before using their works for AI training.
This isn’t just a minor concern – it’s about fundamental transparency in how these systems are built.
When I read about this my first thought wasn’t just about the impressive scale – but about the massive ethical questions this raises.
For a database of this size to be ethically sound, each piece of music used for training should technically be cleared with rights holders. But the reality? This rarely happens.
While the Hualien AI Music Lab’s technology sounds impressive, without transparency about training data sources, it risks perpetuating problematic practices in the AI music industry. This is especially relevant for indie artists building their careers who may have their work used without consent or compensation.
For AI music labs to truly serve communities, they need clear answers to these questions:
- Where does the training data come from?
- Has permission been obtained from rights holders?
- Is there a compensation system for creators whose work is used?
We’ve talked about this before – you can make AI music tools that treat artists fairly. Look at Voice Swap – they compensate singers when users use their voices.
Beatoven is another good example. They’re working with Musical AI to release a song generator in 2025 where all samples are properly licensed. These companies show it’s possible to build AI music tools without taking advantage of artists.
Without addressing these concerns, even the most impressive technology risks undermining the very musical ecosystem it claims to enhance.
The Hualien AI Music Lab has an opportunity to show the right way forward – but only if it acknowledges and addresses these fundamental issues. During my research, I reviewed their website and translated some of the Taiwanese content into English, but couldn’t find any information about where they get their training data from and if they compensate creators for using it.
