Sam Harris recently commented: “We’re never gonna want to see, y’know robots in the NBA.” This claim intuitively seems obviously true but I think deserves more exploration. I don’t think robot basketball is imminent, but the intuitions people use to dismiss it reveal something about the jobs we think of as AGI proof, and why we think that. 

When people are asked about work they think cannot be done by AI, a familiar list of jobs emerges: artists, musicians, teachers, and athletes. Less commonly cited are mechanics, software engineers, and factory workers. The first group involves what people loosely describe as authentic human expression while the second involves technical skill we seem more comfortable thinking of as just computation. 

But this distinction is shakier than it appears. I suspect most people would be perfectly content living in a building designed by AI, provided it’s structurally sound and suited to their needs. Yet these same people would likely report aversion to AI-generated music, even when they can’t distinguish it from human compositions in blind tests.^[1]  There isn’t a consistent standard here, and it’s worth asking whether the discomfort with AI music reflects something deep and principled, or whether it’s closer to status quo bias that will erode as the technology becomes more familiar. It’s relevant that the latter possibility doesn't require believing that there is anything magical about what human musicians do and demonstrates that humans don’t actually have as much of a monopoly on creativity as we would like to think. 

One might worry about the fairness of having robots play in the NBA, competing against humans constrained by their biological limits. But we have to acknowledge that fairness in sports is already pretty incoherent. 

We ban performance enhancing drugs yet allow hypoxic chambers. We heavily scrutinize trans and intersex athletes in female divisions, on the grounds that their elevated testosterone undermines fair competition. And yet we celebrate a 7’1” basketball player whose skeletal structure gives him advantages no training program could replicate. We don’t handicap players for wingspan, VO₂ max, or the genetic luck of fast-twitch muscle distribution. It isn’t logical how we think that winning the genetic lottery is somehow more valid than doping as a route to performance. The difference is that we arbitrarily draw a line between “gifted” and “enhanced”. 

You can’t object to robot athletes on the grounds of their unfair physical advantage when the NBA is already full of humans who won the genetic lottery with their physiology. Maybe the solution is that we only have a single open division, with humans and robots competing together. Objections to something like this on the grounds of fairness are likely harder to defend than it would seem at first. But the problem here is that we’d probably lose interest if robots dominated the NBA, which points us towards what we actually want from sports. 

The most plausible reason robot athletes won’t replace human athletes isn’t fairness, it's that spectator sports are fundamentally about tribal human drama. We’re watching effort, sacrifice, and the failure or triumph of our favourite athlete or team. Most audiences are likely not primarily seeking the highest quality of play when they watch a sports game. A Lakers fan isn’t watching to witness peak basketball, they're watching to see the Lakers win. The tribal identification, built over years, has no obvious analogue with a franchise dominated by robots. 

These observations make me question why we feel connected to certain professionals and not others. One hypothesis for this concerns identification: we connect to athletes, artists, and musicians in part because we’ve attempted some version of what they do. Most people have shot a basketball or done some amount of sketching. We have some sense of the difficulty, and the skills required to be exceptional at these things. This makes elite performance moving in a way that expert plumbing, however technically impressive, usually isn’t. Another is narrative: athletes and artists have stories, achievements, setbacks, and comebacks that we follow over time. Their performance is inseparable from their biography, which is something that robots don’t have. The tension we feel when watching someone whose story we’ve followed is something that will likely be lacking when we watch a robot shoot a three pointer. 

If either of these hypotheses hold, robot athletes face a structural problem unrelated to their skill level. They won’t have coaches who believed in them, injuries that nearly ended their careers, or people cheering them on. The gap between robot performance and human performance might close to zero, or go past it, and it might still not matter, because the audience was never purely watching for the performance itself. 

My tentative view is that Sam Harris is probably right about the NBA, but for reasons more interesting than he suggested. AGI first threatens the domains where the output is separable from the origin. But for sport, art, and music, the human origin is part of the product. The work in these domains will likely be done by humans for a long time to come not because AI lacks some magical creative spark, but because what audiences value in these works is the knowledge that a human made it, struggled with it, and meant something by it. 

These domains might persist not because machines can't match the output, but because matching it isn’t enough. But “might” is still doing a lot of work in that sentence. 

1. ^{^}

   https://newsroom-deezer.com/2025/11/deezer-ipsos-survey-ai-music/