Discover more from Charles Rubenfeld's Newsletter
There’s a bit of a paradox I noticed recently when it comes to live music. If you went to a club that was playing David Guetta music all night, you may not have to pay much to get in. But if David Guetta himself showed up and pressed play on the same music, you may be paying hundreds or thousands of dollars to attend. In the same way, these guys look and sound exactly like The Beatles. But it looks like they play more state fairs than stadiums because they are not actually The Beatles.
In both cases, the output (the music) is essentially the same. But people aren’t just going for the music. The experience of seeing the person perform live is what has the value here. We place value on the “craft” itself, which is why many people are unhappy to find out a performer lip synced a performance even if they couldn’t tell.
There’s been a lot of words written (by me included) about how AI will drive a productivity revolution and transform how every company operates. But predicting AI’s impact on entertainment is as much about predicting how culture will change as it is about efficiency. And based on my observations above, humans (today at least) value human connection which makes AI’s impact on entertainment less straightforward.
Why entertainment is different
When you are paying for a service, I’d argue you generally don’t care that much if a human does it. If you contact customer support, you may prefer a human all things equal but you really just care about getting your issue resolved. If a computer can do it faster and better, great. The same goes for buying stocks, booking flights, or any other service which can be done better and faster without a human involved.
But entertainment is different and I’d put it into two buckets.
Some entertainment is “faceless.” That is, you have no idea who the artist is and have no connection to them. Examples that come to mind for me are:
“Elevator music” you hear in cafes, bars, and shops
Random art you see on the walls of restaurants
Most TV commercials
In principle, I’m sure there are some people who care about whether the jazz music in their coffee shop was created by an AI or whether it’s real actors in the beer commercial on TV. But I just don’t think, on average, people really care that much. I’m betting that Suno and similar AI tools will have a lot of success in replacing this type of entertainment.1
But for a lot of entertainment, people have connections to the artists producing it and care a lot about whether it’s coming from a human or AI. The fact that Taylor Swift is a human with a life behind her music and a story to tell (often at the expense of her ex-boyfriends) matters to people. You may go to see a movie because you like a specific actor, regardless of the plot. So much of fandom today is the act of following these celebrities outside of their actual jobs producing movies, music, TV, etc.
Lessons from the chess world
To look at some of this in practice, consider the Chess world today. AI-powered chess engines became clearly better than humans over a decade ago and since then have become comically good. Magnus Carlsen, the top ranked chess player in the world, would lose 100 out of 100 times if he played the best chess engine today (he doesn’t even bother as a result).
And yet… no one watches AIs play chess today. There was definitely interest in some high profile matches when AIs were surpassing humans’ abilities (ex: Kasparov vs Deep Blue). But today there is basically 0 demand to watch AIs play chess at a much higher level than humans. At the same time, we are seeing an explosion in viewership for humans playing other humans in chess. To use a basketball analogy, people would rather watch humans play high school level basketball against each other than watch AIs play in the NBA.
AI has still had a big impact on the game of chess. Not only do all players use AI powered chess engines as a “copilot” to help them better understand different strategies; these chess engines have fundamentally changed how chess is played.2
AI copilots for stars
I envision something similar for entertainment. The Taylor Swifts of the world will use AI as a copilot to expand their reach, write their songs, and otherwise be way more productive. In many ways this is a continuation of what has already been happening in music. Studios have been using software like LANDR or Izotope for years which uses machine learning in the mixing and mastering process (even if Rick Rubin doesn’t). At the same time, AI will increasingly replace the “faceless” low end of the entertainment spectrum where connection to the human artists doesn’t matter as much.
Copyright will also play an important role in preventing the disruption of stars. If I created an AI generated song with Taylor Swift’s voice, it could go viral3 even if it still wouldn’t be as popular as a real Taylor Swift song (because people know it’s not her). But I can’t do that anyway because I would get sued, just like Suno and Udio are getting sued right now by the record labels. For “faceless” music, I expect copyright will provide fewer protections. If you are creating generic jazz music with AI, who is there to sue you and for what exactly?
AI Singers and Actors?
You might be thinking: what happens if we have AI actors and musicians with looks, personalities and stories just like humans? It's always challenging to predict what the future will look like and whenever you have a technological shift, things that seem crazy end up being totally accepted later on.
That said, I’m generally skeptical and some initial attempts have been clear flops. “Noonoouri” is a digital character with 400k instagram followers and a “record deal” with Warner Brothers, but she seems to have basically 0 listeners to her music.
But the more I think about animated characters, the more I wonder if the idea isn’t so farfetched. Many of us, especially kids, are obsessed with characters like Mario, Homer Simpson, or Frozen’s Elsa. None of these characters are real, but they each have their own backstories and many of us feel connected to them.
Could we see the same with an AI developed singer or actor? My base case is still that it wouldn’t have the same appeal. What does a concert “tour” mean for an AI singer if they can perform at every stadium at once?
But then again, the future is always weirder than we expect!
Subscribe to Charles Rubenfeld's Newsletter
Thoughts on Tech, Finance, and Economics