Can Artificial Intelligence replace Tracy Chapman?

If you kept up with entertainment news last week, you may be aware that the 66th Annual Grammy Awards ceremony took place on Sunday, 4th February 2024. You may also know that one of the highlights of the event was a superb rendition of Tracy Chapman’s classic song, Fast Car, performed by Chapman herself and Luke Combs. The song was originally written and released by Tracy in 1988, and Luke’s 2023 cover has given it a new lease of life, introducing new audiences to an all-time classic while serving a healthy dose of nostalgia to old-timers familiar with the original.

I never thought I’d ever refer to myself as an old-timer, but yes, I do fall into the latter category. Thanks to said nostalgia, I have spent a considerable amount of time in the past week listening to all of Chapman’s music on replay. She’s a songwriting genius, no doubt. Her words and melodies put you in a trance and transport you to a different place and time. Often, it’s place and time that you lived through before, brought sharply to memory by her ever so relatable stories. Sometimes it’s a place and time that you can only imagine, yet she weaves her tales so deftly that you feel like you’ve been there.

You’re probably wondering what all this has got to do with AI, so as the kids say these days, let me land.

Since the advent of generative AI in the past couple of years, there has been a lot of talk about the impact that AI will have on creative industries such as music, movies, television, art, and advertising. From 2nd May to 27th September 2023, the Writer’s Guild of America (WGA) - representing 11,500 screenwriters – went on their second longest strike in history. One of their key objectives in this labour dispute was to obtain commitment from the Alliance of Motion Picture and Television Producers (AMPTP) that AI would only be used to augment, not replace, screenwriters.

More recently, on 27th December 2023, The New York Times filed a copyright lawsuit against OpenAI and Microsoft, claiming that in the process of training their AI models they copied millions of articles from the publication without consent. What aggravates the matter further is the fact that OpenAI’s ChatGPT and Microsoft’s Copilot can generate content that flawlessly mimics The New York Times’ expressive style on request. Regardless of what level of merit you accord to The New York Times’ content and style, it’s tough to argue that the alleged actions of Microsoft and OpenAI do not, at the least, threaten the publication’s ability to protect its content and monetize it.

These are just two of many events that serve to highlight the tension between the creative industry and generative AI. At the core of this tension are three questions:

1.     Can artificial intelligence come up with anything original?

2.     Where do we draw the line between inspiration and copying?

3.     Who owns the output of a generative AI model?

Each of these questions probably deserves an essay of its own, but to illustrate my thinking, allow me to tie this back to Tracy Chapman.

My thinking on this impasse centres on the overarching question of what exactly constitutes good art. The best books I’ve read, the best music I’ve heard, my favourite movies, the best paintings or photographs all have one thing in common – they make me feel something. Listening to Chapman’s At This Point in My Life makes me feel validated, like I am more than the sum of my triumphs and failures. Fast Car makes me feel the heartbreak of shattered dreams. The Promise makes me giddy with the memories of youthful love.

But how much of this feeling is a function of the technical astuteness of the songwriting and musical arrangement? How much of it is a function of the thought that somewhere in the world, singing into a microphone, is a human being who walked my path for a few miles, lived my experience for a few moments, or otherwise gave voice to the deepest recesses of my soul? Can AI generated art really replicate that feeling?

A study led by Luis Arango of the University of Queensland in Australia found that in advertisements for charitable giving, “awareness of the falsity of a face or its status as an AI-generated image has a negative impact on donation intentions”. In other words, empathy is diminished when the protagonist in an advertisement is not human. So maybe the feeling that you get from good art has something to do with the human connection. Maybe. I may yet be proven wrong.

It's inevitable that in the future, someone will use AI to analyse Tracy Chapman’s voice and extract from it just the right mix of frequencies that induce goosebumps. That same someone might also use AI to compose the perfect melody that is bound to stick in your ear for eternity. ChatGPT can already write you a dozen songs a minute in the style of whomever you choose. So, it is inevitable that in the future, I will listen to a song that might move me in the same way that Unsung Psalm does, because it will be a distillation of everything that is good about Chapman’s music.

Will it be an original work of art? Arguably. Will it have been inspired by Chapman or will it be tantamount to downright impersonation? Maybe a mix of both with a leaning towards the latter. Who will own the rights to it? That will depend on how we resolve the two preceding questions. What I can say for sure is that audiences will probably enjoy it a lot less if they find out that it is AI generated.

There is a lot that we do not know about how AI will transform the world. The astronomical valuations of AI companies in the past year are predicated on the promise that it can and will replace a lot of human labour, including that engaged creative endeavours. But how much of that promise is real and how much of it is hype? Big tech has seemingly found a fast car, but is it fast enough so we can fly away?

I’d love to hear your thoughts in the comments below.