"AI Can't Do Creative Work or Novel Discovery"

What do we actually mean by “creative”?

Most of what we call creativity isn’t conjuring something from nothing. It’s taking an idea from one context and applying it to another. A chef borrows a technique from ceramics. An architect steals a structural idea from biology. Newton himself said it best. “If I have seen further, it is by standing on the shoulders of giants.” The giants did the groundwork. Newton made the connection.

This is what humans do. We connect nodes on a vast, expanding graph of ideas. We observe. We experiment. We ask “why?” the same way every three-year-old does, over and over, and each answer opens up more questions. Knowledge isn’t discovered in isolation. It’s emergent. It’s the inevitable outcome of enough observation, enough experimentation, enough connections forming between enough ideas.

And it turns out AI is more than capable of expanding that graph on its own.

In 2024, Google DeepMind’s AlphaFold predicted the 3D structures of virtually all known proteins. That work contributed to a Nobel Prize. Not the Nobel Prize for computing. The Nobel Prize for Chemistry. AI didn’t just assist the discovery. It made the discovery.

Researchers have used AI to identify novel drug candidates that no human team had proposed. AI systems have generated and proven mathematical conjectures that stumped mathematicians for decades. These aren’t parlour tricks. These are genuine expansions of the knowledge sphere, nodes that didn’t exist before.

Now. An important nuance. Not all of these breakthroughs come from the same kind of AI. When most people hear “AI” they think ChatGPT. A conversational language model. But AlphaFold isn’t ChatGPT. Drug discovery models aren’t ChatGPT. The AI landscape is a broad toolkit, and different problems require different tools. Part of what makes this space confusing is that we use one word for a hundred different things. A good advisor helps you figure out which tool fits your problem. A bad one just tells you to plug in ChatGPT and hope for the best.

So. Novel discovery? Settled. AI can do it. Is doing it. Will do more of it.

But what about art?

This is where it gets genuinely interesting, and I want to be honest about the complexity here rather than give you a tidy answer.

I believe the best art is an emotional response to the artist’s unique experience of the world. A song that captures grief in a way that makes you feel understood. A painting that reframes how you see a familiar landscape. A poem that finds language for something you felt but couldn’t articulate. These are acts of emotional translation. The artist felt something real and found a way to make you feel it too.

AI hasn’t felt anything. It has no experience of grief or joy. It has an extraordinary corpus of knowledge about these things, but knowledge about an emotion and the experience of it are very different.

And yet. AI can compose music that sounds good. It can generate visual art that stops you mid-scroll. It can write poetry that reads as competent, sometimes even moving. A lot of people will tell you that AI music is genuinely good. They’re not wrong. They like it. It works for them.

So what’s going on?

Here’s the distinction I keep coming back to. Competency versus relevancy. AI is increasingly competent at artistic production. It can match patterns, combine influences, produce output that hits the right notes (literally). And for a lot of use cases, competency is plenty. Background music. Marketing visuals. First-draft copy. Competent is great. Competent is useful.

But relevancy is something else. Relevancy is the connection between a creator and an audience. It’s the reason a particular song hits you differently than a technically identical one. It’s the reason you care about one artist’s take on heartbreak and not another’s. That connection requires someone on the other end who has actually lived something. Not just processed information about it.

There has always been mediocre art. AI will produce a lot more of it, at scale, and much of it will be perfectly fine for its purpose. The art that genuinely moves people, that changes how they see themselves or the world? That still needs a human in the chain. Not because AI is incapable. Because relevance requires a relationship, and relationships require someone who’s actually there.

Here’s the bottom line for a business leader. AI can already do novel discovery in your domain. It can observe your users, identify patterns you’ve missed, generate hypotheses about product features or market opportunities that deserve testing. That expanding sphere of knowledge? You can accelerate it. Today.

And for the creative work in your business? AI is an extraordinary collaborator. It raises the floor dramatically. But the ceiling still belongs to your people.