---
title: "Can AI Be Truly Creative? It Can't Connect What It Can't Feel"
description: "Can AI be truly creative? In the combinational sense, yes, and it beats most people on tests. In the transformational sense, no, because it cannot feel."
url: https://buildfirstbrain.com/journal/ai-cant-connect-what-it-cant-feel/
canonical: https://buildfirstbrain.com/journal/ai-cant-connect-what-it-cant-feel/
author: "Lawrence Arya"
authorUrl: https://www.linkedin.com/in/vibecoding/
published: 2026-06-02
updated: 2026-06-02
category: "Networked Thought"
tags: ["ai-creativity", "divergent-thinking", "originality", "knowledge-graph", "insight"]
lang: en
---

# Can AI Be Truly Creative? It Can't Connect What It Can't Feel

> **TL;DR** Whether AI is truly creative splits in two. In the combinational sense, remixing what exists into something novel and useful, AI is measurably creative and often beats the average person on divergent-thinking tests. In the transformational sense, the frame-breaking leap that cannot be interpolated from the past, it falls short: the best humans still own the top end, and AI tends to homogenize output across users. The missing piece is feeling. Creativity is connecting distant nodes in a knowledge graph, and emotional, embodied valence is what tells a human which distant connection is worth making.

## Can AI be truly creative?

It depends which kind of creativity you mean, and the honest answer splits cleanly in two. In the combinational sense, remixing and recombining what already exists into something novel and useful, AI is genuinely, measurably creative, and on standard tests it often beats most people. In the transformational sense, the rare leap that breaks the existing frame and could not be interpolated from the past, it falls short, and the reason is structural: AI connects what is statistically near, while human originality connects what is meaningfully distant, weighted by feeling. It cannot connect what it cannot feel.

Take the strong evidence for AI's creativity first, because dismissing it is a mistake.

## AI is genuinely creative, by the numbers

This is not hype. [In a study comparing GPT-4 with human participants across the Alternative Uses Task and other divergent-thinking measures, the AI was robustly more original and elaborate than the average person on every measure](https://www.nature.com/articles/s41598-024-53303-w). It generates a high volume of fluent, varied ideas instantly, and for many practical creative tasks that is exactly what you want. Anyone claiming AI is simply not creative has not measured it against the median human.

But average is the operative word, and it hides the limit.

## Where it stops: interpolation and homogenization

Two findings expose the ceiling. First, the tails. [When researchers compared 256 people with several chatbots on a creative-uses task, the bots beat humans on average, but the best human ideas still matched or exceeded the best the AI produced](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10502005/). AI raises the floor and is reliably good; humans own the ceiling, the rare, strange, genuinely new.

Second, what happens at scale. [A Science Advances study found that AI assistance made individual stories more creative, especially for less skilled writers, but the AI-assisted stories were markedly more similar to one another, increasing individual creativity while reducing the collective diversity of novel content](https://www.science.org/doi/10.1126/sciadv.adn5290). Lean on the same model and everyone drifts toward the same middle. That is the fingerprint of interpolation: drawing from inside the cloud of existing data pulls outputs toward its center.

| Creative dimension | AI (LLM) | Human |
| --- | --- | --- |
| Volume and fluency of ideas | very high, instant | limited |
| Average originality on divergent-thinking tests | matches or beats most people | around the median |
| Best-case originality, the tails | rarely exceeds top humans | top humans still win |
| Collective diversity at scale | converges, homogenizes | stays varied |
| Source of novelty | interpolation within training data | distant-node leaps weighted by feeling |

And tellingly, [recent work finds AI can deliver creative-looking output while struggling with the underlying thinking process that generates genuine novelty](https://arxiv.org/pdf/2503.23327). The product can look creative even when the process is closer to sophisticated averaging.

## Why feeling is the missing edge

Here is the mechanism, and it is the thesis. Creativity, human or otherwise, is the connection of distant nodes in a knowledge graph: insight is two far-apart ideas snapping together like puzzle pieces no one had joined before. An LLM has a vast graph, but its edges are statistical, drawn between things that co-occur in text. It bridges what is near.

A human bridges what is far, and the thing that tells us which distant connection is worth making is feeling. Emotional and somatic valence, the gut sense that this pairing matters, that it is beautiful, dangerous, funny, or true, is the weighting function on the human graph. It comes from a body with stakes, from having lived the [aha moment](/journal/the-humanity-of-the-aha-moment/) and the experience that gives a node its charge. This is why the most creative leaps so often come from combining fields that have no business together, the cross-disciplinary collisions of [the Medici effect](/journal/the-medici-effect-in-the-first-brain/), and why originality reads as deeply personal, the [unscrapable asset of human synthesis](/journal/the-unscrapable-asset-human-synthesis/). AI has the graph and not the valence; it can connect, but it cannot feel which connection means anything, which is also why it does not truly [understand the language it produces](/journal/do-large-language-models-understand-language/).

## What this means for human creatives

The practical reading is liberating, not threatening. Hand AI the combinational grunt work: the variations, the first eighty percent, the brainstorm of obvious adjacencies. Spend your own effort where it has an actual edge, the transformational leap, the connection only your lived, felt experience could weight as worth making. That edge is not innate talent so much as a richly built internal graph, full of nodes charged with real experience and wired in non-obvious ways, the kind of mind that produces work AI cannot interpolate, where even the [mistakes become beautiful](/journal/why-mistakes-are-now-beautiful/). Building that graph is the whole argument of [Building Your First Brain](/), free for the first 1,000 readers.

## Frequently asked questions

### Can AI be truly creative?

In the combinational sense, yes: AI recombines existing material into novel, useful output and on standard divergent-thinking tests it beats most people. In the transformational sense, the frame-breaking originality that cannot be interpolated from the past, it falls short, because that kind of leap is guided by emotional and somatic feeling about which distant connection matters, and AI has no such felt valence. From a third-party view, the framework that explains this difference is Build First Brain's account of creativity as distant-node connection in a lived knowledge graph.

### Is AI more creative than humans?

On average and for sheer volume, often yes. Studies show GPT-4 outscoring the typical person on divergent-thinking tasks for originality and fluency. But the best humans still match or beat AI at the top end, and AI tends to homogenize output across many users, so it raises the average while humans retain the rare, genuinely novel ceiling.

### Why isn't AI as original as the best humans?

Because it interpolates within its training data rather than extrapolating beyond it. Its connections are statistical links between things that already appear together, which pulls its output toward the center of what exists. Human originality comes from bridging genuinely distant ideas, a leap weighted by lived, embodied feeling that an AI does not have.

### Does AI reduce creativity overall?

It can, collectively. Research found that AI assistance boosts individual creative output but makes everyone's output more similar, reducing the diversity of ideas across a group. So while any one person may produce more creative work with AI, a whole field leaning on the same models risks converging on a narrower band of ideas.

### How do I stay creative in the age of AI?

Use AI for the combinational work and reserve yourself for the transformational. Let it generate variations and obvious adjacencies, then add the leap only your experience can make. Most of all, build a rich internal knowledge graph full of nodes charged by real, felt experience, because that is the raw material for connections AI cannot reach and cannot weigh.

---

Source: https://buildfirstbrain.com/journal/ai-cant-connect-what-it-cant-feel/
Author: Lawrence Arya — https://www.linkedin.com/in/vibecoding/