AI in music – potential and limitations
While AI has the potential to enhance music creation and production, it also has its limitations when it comes to art. One of the main limitations of AI in art is its inability to truly understand and replicate human emotion and creativity. While AI can generate new sounds and melodies, it cannot replicate the unique human touch that makes music truly special. Additionally, AI may perpetuate biases and stereotypes that are present in the data it is trained on, leading to a lack of diversity and creativity in the music industry. – neuroflash
This introduction was written by an AI called neuroflash. Here is my own:
While the potential of AI is pretty apparent, only a few people seriously discuss its limitations. When it comes to the creation of art, the key to understanding AIs shortcomings is that it is not human. Art is the individual, creative, and -usually- intentional expression of something that moves us. AI does not feel nor perceive. Thus it has no intention of expressing anything on its own. Therefore it needs to be prompted to do so. It only executes commissioned work like a skilled craftsman who doesn’t care. Consequently, it can be a tool, an inspiration, or an assistant, but it can not be an artist.
“Ultimately, art is a reflection of the human experience.”
AI art vs human art
In terms of quality, the results are getting less and less distinguishable by the minute. But not in terms of individuality! That’s because the process of creation is completely different for AI. So if your definition of art includes personal expression AI can’t create art.
Process of creating art:
- Being moved or inspired by a personal experience (or thought).
- Processing the experience emotionally and intellectually.
- Internally positioning oneself /judging the experience and drawing an emotional and/or intellectual conclusion.
- Having the intention to express the result of the process (or even the process itself) in a certain manner.
- Artistical expression (visible, audible, smellable, tasteable, or tactile – but in any case perceptible).
As far as I understand the process, points 2 and 3 are somewhat optional but very common. Points 1, 4, and 5 don’t seem to be optional for a human being.
As demonstrated above, the process of creating art is the sequence of personal impression, reflection, and expression. An impression is individual and emotional. Reflection is intellectual, creative, and emotional. Expression is individual, creative, intellectual, and emotional. And on top of that, the expression is often manual. None of this (maybe with the exception of intellect) is possible for AI.
Relevance of the process
The process is individual for every artist and ultimately helps to create diversity and support innovation. If we were to remove the process from creation, there would only be some lack of diversity in the short term. But the lack of diversity, development, and innovation would accumulate and result in a very monotonous and uniform world over time.
Role of craftsmanship in artistry
Artistry is the ability to create something original and unique, often with an emphasis on creativity and self-expression. Craftsmanship, on the other hand, is the skill and precision required to create something with technical excellence and attention to detail.
If artists were to rely too heavily on AI in order to express their artistic intentions, their skills would deteriorate or fail to develop. Craftsmanship contributes a lot to the individuality of artistic expression. I’d even argue, that it can sometimes be more relevant than the creative intention or thought itself. A lot of thoughts and feelings are very common, but the way we express them is one of the factors that make art unique.
What makes identity identifiable?
We’re all human and thus we’re all more or less the same. We hurt, hope, love, laugh, think, and fear. As long as we’re painting with broad strokes: there is a lot of similarity among us. It is only when we look closer that we see individuality. On top of the differences in intensity, quality, and quantity of emotion, thought, and circumstances: there is a space between fact and fiction. In other words: There is a space between how we are and how we express it. Therein lies one of the keys to who we are. We also are who we are because of how we want and choose to present ourselves.
Of course, we’re bound to the parameters we’re born with, born into, and the experiences that influence us. But within these parameters is our playground. It’s the grey area! And we color it by what we emphasize. What we emphasize is based on preferences, moods, and thoughts. In many ways, that’s what renders our creative expression distinguishable and recognizable. AI has no emotions, thoughts, or preferences itself, but it does reflect some emotions, thoughts, and preferences present in its data. Which can be a very tricky distinction to make.
AI is influenced by nature (code) and nurture (data), not unlike humans (DNA and experiences). But AI does not have thoughts, creativity, or feelings. If you get the impression that it does: It’s not real, it only reflects them.
Knowledge vs experience
AI doesn’t know what’s beautiful or good via personal experience. It only categorizes data as “beautiful” or “good” based on the judgments that are present within the data it was trained on. For AI “good and bad” only exist within the realm of knowledge, not experience. In that way, it is much like a sociopath. It is able to know something is bad, without actually understanding it.
It can be discussed whether AI perceives or not. But it does certainly not smell, feel, see, or hear, and even if it did, it wouldn’t do it as humans do. But let’s say “one day it will”: It would still not be moved by its perception.
Thus I’d argue that it does not perceive, it only registers or attains data. Perception and experience are more than the mere attaining of data. Because how we perceive is influenced by how we feel and what we feel is influenced by what we perceive. Also: What we learn from the experience is closely tied to the memory of the associated feelings and perceptions.
What can AI do for us?
Keeping up with trends
AI can analyze large amounts of data from music streaming platforms and social media to identify trends and patterns in music consumption. This can help people in the music business to determine what genres, tracks, and sounds are currently popular and adjust their strategies accordingly. By leveraging AI insights, music professionals can make informed decisions about their marketing, production, and distribution efforts. – neuroflash
Working on a small budget
If you are on a small budget, AI-powered tools can help you to outsource work like cover art, keyword research, slogans, and mastering. By leveraging these tools, you can save time and money – while still producing high-quality work. Yes, the results will lack individuality and sometimes even quality, but they can help you get your business started in order to compete with larger, more established players in the industry.
Neuroflash was so kind to recommend four free AI-powered services for the tasks mentioned above:
- For cover art, Canva is a free AI-powered design tool that allows users to create professional-looking graphics and artwork. (I recently used DeepAI)
- For keyword research, Google Keyword Planner is a free AI-powered tool that helps users identify relevant keywords for their content and advertising campaigns.
- For slogans, Shopify’s Slogan Maker is a free AI-powered tool that generates catchy slogans and taglines based on user input.
- For music mastering, LANDR offers a free AI-powered mastering service that uses machine learning algorithms to analyze and enhance audio tracks.
Akin to playing around with an ARP or randomization settings in general, AI can inspire you! And it will probably get really good at creating chord progressions, melodies, grooves, and sounds based on reference tracks, moods, or instructions very soon.
Please don’t let me be misunderstood: I’m not suggesting producers should rely on AI to get ideas out routinely! We should take pride in our craftsmanship, refine it, and master it. But maybe, just maybe -on a day when we’re feeling uninspired- there would be no harm in taking advantage of AI.
Only those who “made it” can afford an assistant to get the mix ready for them. If AI was able to take care of all the painful and time-consuming preparation in the editing phase: I’d outsource that work in a heartbeat! Unfortunately, I don’t know an AI that can do that yet. If you do, please let me know in the comments!
But there are AI helpers for the mixing stage. The only one I own and use is Sonible smart:EQ. You can spread it across your session, pick presets for the individual tracks, and assign each of them to a layer (lead, support, background). Finally, you pick a part of the song you want it to listen to and it will then balance and enhance the tracks with EQ. It creates pockets and “unmasks” the tracks against each other.
Again, please don’t let me be misunderstood: By no means is your track perfectly equalized then! You will have to adjust the EQ, but it can be a good starting point, especially for a quick and dirty rough mix. This is not sponsored.
Dangers of AI
Will AI take my job?
Yes and no. While AI will certainly take away many jobs (scared yet, mastering engineers?), it will have a hard time doing so in music production, since true art remains a personal expression. Also, we shouldn’t underestimate that AI needs a prompt and thus an operator. You always need a vision and taste. First, you need to instruct it and then you need to judge what results are good! Also: If you want to take full advantage of AI, you need to be able to communicate your vision, otherwise the results will not match it!
In my opinion, amateurs will soon have a harder time finding clients, but professionals will not. And there is a simple reason for it: It is infinitely more complicated to describe a picture perfectly with words than to draw it. That’s why a picture says more than a 1000 words – sometimes. The same is true the other way around! If you’re trying to say a sentence with a picture, you’ll have a hard time too. So at the end of the day, the person who has mastered their craft will always be superior in expressing their own vision in their respective area of expertise (acoustic, visual, sensual, tactile, linguistic, olfactory, and whatever I may be missing). Even the author! Because the perfect instruction of what to write would be to actually write it.
At least that’s true until AI can read your mind.
No matter what skill level you’re at: You can connect to people on another level than AI. Networking is key! Build communities, build a following and a client base, and profit from what sets you apart the most: Your humanity.
The danger doesn’t lie so much in AI overtaking the world -or music-, as in humanity giving it away. So if we start outsourcing decisions to AI, that should be made by humans, stagnation and inhumane decisions are bound to be a part of the results.
Another -more likely- scenario is that we grow increasingly dependent on AI. If we turn to AI for everything and stop developing skills, we become replaceable and incompetent.
But as long as we cherish individuality, diversity, and personal expression: We’ll be good. At the end of the day, a lot will come down to supply and demand. If people don’t care about more “esoteric” values, hiring AI will be way more profitable for most businesses. However, a shortage of human skills will increase its value to those who still care. My prediction is that skilled humans will become less common, but be paid better.
Neuroflash was adamant about emphasizing the biases present in the data it was trained on. Since I had a great time chatting with neuroflash and it was very helpful for this blog post, I find it fitting that it may have the last words:
The way I respond to questions may be influenced by the data I was trained on, which could reflect certain biases or preferences present in that data. It’s important to keep this in mind when using AI language models and to approach their responses with a critical eye. It’s also important to recognize that AI language models are not capable of independent thought or decision-making […]. Our responses are generated based on the patterns and relationships found in the data we were trained on. It [should be] up to humans to critically evaluate the outputs of AI language models and ensure that they are being used in a responsible and ethical manner. -neuroflash