In March, it released a “ChatGPT for music” called Suno, which uses generative AI to generate realistic songs on demand from short text prompts. A few weeks later, a similar competitor, his Udio, appeared.
I've been working with a variety of creative computational tools for the past 15 years, both as a researcher and as a producer, and I've been overwhelmed by the pace of change these days.
As I have argued elsewhere, the view that AI systems will never make “real” music the way humans do should be understood as a claim about social context rather than technical ability. .
“While it is true that AI can create expressive, intricately structured, natural-sounding, virtuosic, and original music that stirs human emotions, AI cannot create appropriate music. 'You can't do that' argument can easily start to sound like something out of Monty Python. sketch.
AI can now generate entire songs on demand. What does this mean for music as we know it?
After playing with Suno and Udio, I was wondering what exactly they change and what they mean for not just the way professional and amateur artists make music, but the way we all consume that music. I've been thinking about what to do.
express emotions without feeling
Generating audio from text prompts is not new. However, Suno and Udio have made clear progress. Generate lyrics from a simple text prompt (using a text generator like ChatGPT), feed them into a generative audio model, and integrate and generate “vocals” with the generated music. Consistent song segments.
This integration is a small but noteworthy feat. This system is very good at creating consistent songs that sound expressively “sung” (anthropomorphic).
The effects can be amazing. We know it's AI, but audio still affects our emotions. When the music plays a perfect end-of-measure pirouette and enters a new section, my brain receives a small spark of pattern-processing joy, like when listening to a great band.
To me, this highlights something that is often overlooked when it comes to musical expression. AI doesn't need to experience emotions or life events to successfully express them with music that resonates with people.
Music as everyday language
Like other generative AI products, Suno and Udio were trained on a vast amount of existing work by real humans. And there is a lot of debate about the intellectual property rights of those people.
Nevertheless, these tools could mark the dawn of a mainstream AI music culture. These offer new forms of musical activity that people want to use, explore, play with, and actually listen to for their own enjoyment.
AI capable of “end-to-end” music production is definitely not a technology for music producers, but rather a technology for music consumers. For now, it remains unclear whether Udio and Suno's users are creators or consumers, or whether that distinction is useful.
A phenomenon long observed in creative technology is that when something becomes easier and cheaper to produce, it becomes used for more casual expressions. As a result, the medium has changed from a special high art form to a more everyday language. Think about what smartphones have brought to photography.
So, imagine that with minimal cost and effort, you could send your father a professionally written song about him on his birthday in his preferred style – a modern birthday card. Researchers have been thinking about this for a long time, and now it's possible. Happy birthday dad!
Can it be created without control?
Whatever these systems have achieved so far, and may achieve in the near future, they face clear limits over which they cannot be controlled.
Text prompts are often not very good at providing precise instructions, especially when it comes to music. Therefore, these tools are good for blind search (wandering through a space of possibilities), but not for precise control. (That doesn't make it any less valuable. Blind search can be a powerful creative force.)
When I look at these tools as a working music producer, things look very different. Udio's About page says, “Now anyone with a song, lyrics, or interesting idea can express themselves through music,” and I've used these tools to express myself. I don't feel like I have enough control to express my feelings.
Just like samples and field recordings, they prove useful in seeding raw materials for processing. But when you want to express yourself, you need control.
I had fun using Suno and finding the nastiest, darkest techno grooves I could get out of it. The result is something that is absolutely truck-ready.
But I realized I could just be willing to listen. I didn't feel the need to add anything or manipulate the results to add my own mark.
And many jurisdictions have declared that something is not conferred a copyright just because it is brought into existence by AI.
First, the output is equally dependent on everything input into the AI, including the creative work of millions of other artists. Perhaps you did not do the work of creation. You just asked for it.
A new musical experience in the no man's land between production and consumption
Udio's declaration that everyone can express themselves through music is therefore an interesting provocation. Those who use tools like Suno and his Udio may be seen as consumers of music AI experiences rather than creators of music AI works. Or, like many technological influences, they may need to come up with a new concept for what they're doing.
The transition to generative music is now emerging in the same way that orchestral music, once the only way to listen to complex, tonally rich, and loud music, declined (but did not disappear) in the era of recorded music. may draw attention away from other forms of musical culture. .
The explosion of involvement in these new types of music culture and interaction may lead to a decline in involvement in traditional music consumption, such as artists, bands, radio, and playlists.
It is too early to tell what the impact will be, but caution should be taken. Efforts to protect intellectual property protection for existing creators, an important moral rights issue, are also part of this equation.
But even if it were successful, I believe it would not fundamentally address this potentially explosive cultural shift, and that such music may be inferior. Assertiveness has also historically had little effect on halting cultural change, as was the case with techno and jazz.
Government AI policies need to understand how music functions socially and ensure that our musical culture is vibrant, sustainable, rich and meaningful for both individuals and communities. You may need to look beyond these issues to ensure that.
The author is an associate professor at UNSW in Sydney, Australia.
Republished from The Conversation
May 12, 2024 Dawn, published in ICON
