Let's settle the score on AI music (and other art)!
A controversial point of view for now, but I feel like it'll become mainstream in no time!
đ THE POINT IS: If you outsource your thinking and your agency to AI, then yes, you will most likely get slop. But if you put your soul onto the canvas and work with an AI voice, or band, or hand, or producer to get a meaningful representation of something youâd never be able to do on your own otherwise, then thatâs just another form of art.
Click the âheartâ button above to let me know how you feel!
I am an AI artist.
Phew! There. I said it!
Iâve used tools like Donna AI and Suno to create music from lyrics that I wrote. Actually, those lyrics started off as poems that I wrote over the past decade or so. I got the idea one night to try to take a poem of mine and throw it into Donna AI to see what it would do with it.
I quickly learned that that doesnât work.
I went back to the drawing board to turn the poem into a set of musical lyrics. I added instrumental and vocal cues to direct the digital performers on how to create the music. Then after probably 30-50 iterations, tweaks, changes, suggestions, and instructions, for the vocalist and the band, I finally heard a version of the song that sounded like, well a real track (by the way, for anyone who has been in a vocal or instrumental group, I hope I didnât just give you PTSD!).
The point is that it was my work, my labor, my idea, my intellectual property, my directions, my ears, my brain, and my soul that went into all of thatâŚjust like any other songwriter in the world. Why is âsongwriterâ in italics back there? Simply because most songwriters donât play a band worth of instruments. Many donât singâŚwhich is why they work with a vocal artist and a musical group to ultimately bring their vision to reality.
Be sure to click the âheartâ on this article and if youâre not already:
To me this kind of work is uncovering the difference between AI slop and human-with-AI created art.
I can hear it now: âThatâs just not the way it goes! AI canât create art even if your hand is on the brush.â
Letâs take a walk back in time, shall we?
The year is 1982 and this brand new technology is hitting music stages around the world. Itâs electronic and odd and makes all kinds of sounds that arenât real. The synthesizer was so much as banned by the UK Musicianâs Union in order to protect jobs. Queen had a âNo Synthesizers!â advertising campaign because they were soulless and didnât create ârealâ music. By 1988, though, Mannheim Steamroller was releasing one of their famous holiday albums. I was listening to âCarol of the Bells,â on my commute this morning thinking about how powerful that song is and it's completely made with synths.
How about the rise of Auto-Tuners in the 90âs? At one point these handy little tools helped studios configure their equipment to align to the artistâs voice. However artists like Cher and Kanye West started using these tools to distort their voices on purpose, thus making it possible to create new tones and to express their souls more completely in their works.
As we look at musical history, we can see many examples of where a new instrument or tool came onto the scene and met instant rejection by existing artists who felt like their work was more ârealâ than the artists using these new technologies. We havenât even talked about the drum machine âpanicâ in the early 80âs, but I think you catch the beat đ
By the way this is true for other art formsâŚdid you know that when cameras came on the scene, painters said that photography wasnât real art because âthe machine does the workâ! In 1859, the famous poet and critic Charles Baudelaire wrote that photography was
âartâs most mortal enemyâ and a refuge for âpainters too lazy to complete their studies.â
He sounded exactly like Nick Cave does today.
Yeah but I mean there is such thing as AI slopâŚright?
Absolutely!
I think the problem with AI in art is the same as the problem with AI in just about everything else: if humans give their agency away, let AI create an artifact, and donât take ownership of the final product, we often see slop arise. Iâve definitely just told Donna to create a silly song in some style about my dog and let it rip. Now I actually like that song about my dogâŚbut thatâs not the point. The song isnât going to win any awards and if I put it online, itâd get ripped to shreds. But thatâs the point: I was playing around, maybe getting some ideas, learning how to use the tool, and I wasnât trying to pass it off as Beethoven.
Nick Cave came down hard on a fan who asked ChatGPT to generate lyrics in Nickâs voice. He said
âData doesnât suffer. ChatGPT has no inner being, it has been nowhere, it has endured nothing... This song is bullshit, a grotesque mockery of what it is to be human.â
âŚand well, heâs right! In my example above, I wrote the lyrics. I translated my lived-experiences into a workable, rhythmically acceptable (and dare I say enjoyable) set of words that convey deep meaning. I did not give my agency away. I did not give my creativity away. In fact it took days to finesse the work into something that sounds somewhat professional. Just like the early photographers who were told that they donât draw every line of their photographs so they canât be called art, I didnât write every note of those songs, but I framed up the emotions and instructed my tool to help take that out of my soul and into this reality.
Be sure to click the âheartâ on this article and if youâre not already:
This distinction between outsourcing agency (slop) and directing outcomes (art) doesnât just apply to Suno; it also applies to your GitHub repositories.
Itâs always good to make sure weâre lifting our heads up to hear whatâs happening in the AI world (or otherwise!) in order to build perspective. This is, at its core, a discussion about how AI tools should be used in our lives.
Anthropic recently said that 90% of the code being written is done by Claude Code. The role of the software engineer is fundamentally changing to a supervisor and a guide over someone who puts fingers to keyboard to belt out a few thousand lines of instructions. Is that wrong? Is it AI slop thatâs being created? Are the humans doing just as much work reviewing the code and making sure it works?
These are interesting questions, maybe, because major vibe-coding platforms, such as Cursor, have their own test beds and automation tools built into the engines too. In fact, one of the earlier ways that AI tools for coding (like GitHub Co-pilot) were used was to create test cases and test scripts based on code (in many cases on older code or code that was being updated for a release). So why not just let these babies rip and do all of the work? What do you mean 90%? Why isnât it 100%?? Will the software engineer truly go the way of the Dodo by next year?
Clearly weâre not at 100% and frankly when I actually put my hands on the keyboard to vibe code, many times it doesnât even run. So to me there is clearly hype out there; clearly weâre somewhere more realistically around 50% of code generated; that code may or may not be actual, production-grade, enterprise hardened code (probably not); and in fact thereâs a much longer tail that we have to look forward to. This isnât just speculation. Last week during the AI Daily Brief we heard about user stats starting to slow on platforms such as Cursor and Replit because the whole vibe-coding âthingâ isnât going as smoothly as everyone thought. A recent study by METR (Model Evaluation & Threat Research) found that AI coding tools actually slowed down experienced developers by ~19% because they spent so much time reviewing and fixing the code. Weâre seeing the âProductivity Paradoxâ emergeâŚhumans canât just watch the loop. They need to be in the loop. Thatâs not to say that code wonât be created by AI-developers, but much like my music example above, the better strategy for now anyway is still to use AI tools as thought partners and let them be âtandem codersâ with seasoned software engineers.
Thought partner you say? Thatâs a new oneâŚwhat do you mean by that?
Iâd hate to steal the thunder from Geoff Woods in his book, the Ai-Driven Leader, because that book is a must read for anyone using AI tools for basically any reason. In that book, though, Geoff presents many eye-opening and simple concepts on how to use AI tools to truly assist you, the thought leader. Iâve tried implementing a few of the prompts that he provides in the book and seeing how ChatGPT and Gemini behave differently when theyâre put into this âframe of mindâ is nothing short of mind blowing.
We are moving quicklyâŚ
âŚinto an era where our everyday toolset is transforming right before our eyes. Whether itâs the new âcameraâ for current day photographers or the âsynthesizerâ and âdrum machineâ of the mid-20âs for todayâs audio artists, new tools will define new techniques, genres, and outputs. AI slop is realâŚbut itâs not fair to throw that label out to everything that AI touches. Humans will create works with AI tools just like they started creating new vocal intonations with auto-tuners.
In the workplace, weâll see people succeed by working with AI tools, using them as tools and not expecting that theyâll take away all of the value that workers provide. Donât get me started on my discussion about the future of workâŚthatâs for another article. Suffice it to say, though, the faster we learn how to think 10x differently, the smoother weâll move into the AI-powered futureâŚwith human hands hopefully on the wheel!