Why AI Won’t Replace Songwriters Any Time Soon
With the emergence of AI songwriting platforms like Sono and Udio, there's been a lot of speculation about AI replacing songwriters. Here are three reasons that's not going to happen soon.
AI aggregates the most probable. When it comes to words, AI is set to complete thoughts in the most logical and probable fashion. This works great for expository writing like reports or studies but lacks the unexpectedness that makes hit songs hit songs. Sure, you can ask AI to write "in the style of..." but it will be aggregating and synthesizing ideas from its analysis of existing sources, not coming up with completely fresh ideas. You can invite AI to write "in the style of..." Taylor Swift, but Taylor Swift is going to put her on spin on whatever she comes up with - something that AI at present can only replicate.
AI can learn, but it can't observe and blend observations in a human way, filtered through lived experience. AI can put ideas together from many sources, but it lacks the power (so far) of human sentience, consciousness and observation across a wide range of scenarios - the stuff of inspiration that makes for fascinating songs. Imagine trying to have AI write an original song like Eleanor Rigby or Penny Lane. AI can imitate the style, even make up a narrative, but it won't be from AI's observations. It will be a facsimile, algorithm or replication, not an original work observed or imagined from lived experience. For example, it may know that tears often follow sadness. But that’s the same as crying because you feel sad.
AI can't (as yet) think well multi-dimensionally in the ways humans do. Part of what makes human creativity surprising, novel and intriguing is the way songwriters pull from a wide variety of sources and use words in new ways or with alternate meanings. AI is not set up to “think” that way. The way its inventors have programmed it so far is to think linearly and logically. Think of a song like All About The Bass. Equating "bass" with "booty" and then understanding why that might be titillating to humans, is something that, so far, AI can’t replicate, since it has no lived experience, cannot "feel" nor come up with feelings on its own. It can only replicate human emotion based on millions of samples.
Likewise, with music composition, AI functions a lot like autocorrect in typing. It can come up with chords harmony and melody that "makes sense" from analyzing millions of songs. But it's not making anything wholly original, just re-combining elements based on what it's sampled, analyzed and what's most "probable" to replicate whatever prompt its given. But it lacks human ingenuity built on intuition and the feelings that musical elements evoke in humans. As for now, it has no real understanding of the "why" part of music.
Which is not to say that people won't use AI to generate musical or lyrical ideas to develop. However, it still takes a human to add the human touch, intuition, observation and surprising originality.
Will people try to have AI write a "hit" song? Sure, people will try and maybe one of those tries will become a hit. It could happen, but the interplay of thoughts and human emotions and the ideas that evoke them are mostly outside the grasp of AI except as exercises in recombining already existing material.
So keep doing what you're doing, all you songwriters and composers and producers. The listening public still needs your uniquely human innovation and perspective. As for now, AI remains a tool to explore, but not a replacement for actual artists.