# Unit 12: What's Yours

> From 101 – Fundamentals on ThinkModel. Free to read; the interactive version of this unit is at https://thinkmodel.ai/course/101/unit/12

*Last updated: May 31, 2026. Statistics on AI economics and creative work were updated to current figures.*

## The Hook

AI can write a solid essay in 10 seconds. It can generate album art that looks like it took weeks. It can compose music, write code, draft business plans, and create marketing copy that passes for professional work.

So here's the question nobody is asking loudly enough: if AI can do all of that, what's the point of you?

This isn't a rhetorical question. It's the real one. And the answer isn't "don't worry, AI will never replace humans" — because in some tasks, it already has. The answer is something more interesting and more demanding than that.

**Video:** The Origination Divide

## The Core Concept

A study published in January 2026 by researchers at the Université de Montréal tested AI against 100,000 humans on creativity tasks. The results: AI beat the average human, outscoring roughly 72% of the people tested on measures of creative thinking.

But — and this is the important part — the top 10% of creative humans still far exceeded AI. The most creative people weren't just slightly better. They were in a different league. AI could produce competent, fluent, even surprising creative work. What it couldn't do was originate a perspective, a worldview, a voice that comes from having lived a specific life.

## Case study: The Montréal creativity study: What AI can't originate
The January 2026 study from the Université de Montréal was the largest ever comparison of human and AI creativity. Researchers administered standardized creativity tests, built around the Divergent Association Task, to more than 100,000 human participants and several frontier AI models.

AI outperformed the average human on most measures. On divergent thinking tasks, GPT-4 beat about 72% of human participants — strong, but not the near-perfect score it's sometimes reported as.

But the top human creators demolished AI on tasks requiring origination — original metaphors drawn from personal experience, novel problem frames that challenged the premise of the question, creative solutions that required understanding real-world context the AI lacked. AI is excellent at recombination. Humans are capable of genuine origination.

This is the key distinction: **AI can produce, but it can't originate.**

It can generate a poem in the style of any poet who ever lived. It can't write a poem that comes from your specific experience of watching your grandmother garden on a hot summer day. It can write a business plan based on patterns from thousands of successful companies. It can't write one driven by your specific insight about a problem nobody else has noticed.

Production is about fluency. Origination is about having something to say.

The students who thrive in an AI-saturated world won't be the ones who outsource everything to AI. They'll be the ones who know what they think *before* they ask AI to help them say it. Who have a perspective, a taste, a point of view that AI can amplify but never replace.

This isn't warm-and-fuzzy reassurance. It's practical advice. If your only skill is producing competent text, images, or code — AI does that cheaper and faster. If your skill is knowing *what to produce and why* — what matters, what's missing, what people actually need — that's the skill AI makes more valuable, not less.

**Check your understanding.** The Montréal creativity study found that AI beat the average human on creativity tests. What does this actually mean for human value?

- A. Humans are no longer needed for creative work, since AI now scores higher
- B. AI is more creative than every human, since it beat the people tested
- C. Producing competent work is now cheap, but only people can originate
- D. Creativity tests miss real creativity, so the finding tells us nothing

Now zoom out from the individual to the system. AI doesn't just raise questions about personal identity. It raises questions about power.

The AI industry is remarkably concentrated. Private AI investment is dominated by a few countries: $285.9 billion in the US in 2025, $12.4 billion in China, and sharp drop-offs after that. The companies building frontier models — OpenAI, Anthropic, Google DeepMind, Meta — are backed by the largest corporations on Earth. The hardware comes almost entirely from one company (Nvidia). The talent pool is concentrated in a handful of cities.

This means the values, priorities, and blind spots of a very small number of people are shaping a technology that will affect everyone. What languages AI speaks best. What cultural contexts it understands. What it refuses to do and what it happily does. What gets funded and what doesn't.

## Deep dive: The concentration of AI power
The numbers paint a stark picture. In 2025, global private AI investment reached roughly $345 billion — and $285.9 billion of that went to US-based companies, more than 23 times China's $12.4 billion. Other countries trailed far behind.

Hardware concentration is even more extreme. Nvidia controls approximately 80% of the AI chip market. A single company's product roadmap effectively determines the pace of AI development worldwide. When Nvidia can't produce enough chips, AI progress slows. When it releases a new architecture, capabilities jump.

Talent concentration mirrors investment: the majority of frontier AI researchers are clustered in a few cities (San Francisco, London, Seattle, Beijing). The decisions about what AI can do, what it refuses to do, and whose needs it prioritizes are being made by a remarkably small, remarkably homogeneous group of people.

The AI Act — the world's first comprehensive AI regulation — entered force in August 2024. The US took the opposite path: President Biden issued a broad Executive Order on AI safety in October 2023, but President Trump revoked it on his first day back in office in January 2025 and replaced it with one aimed at removing barriers to AI development. There is still no comprehensive federal AI law. Meanwhile the states moved on their own — 131 state-level AI laws passed in 2024 alone, nearly triple the 49 passed the year before, with more every year since. Governments are trying to figure out the rules. But the technology is moving faster than the regulation.

Meanwhile, the question of who owns creative work is headed to the courts. Artists have sued Midjourney and Stability AI for training on their work without permission. Disney and Universal sued Midjourney for reproducing copyrighted characters. The US Supreme Court declined to hear a case arguing AI-generated art should be copyrightable — effectively reinforcing that copyright requires human authorship.

## Warning: Common misconception: AI-generated content is yours to use however you want
The legal landscape for AI-generated content is actively being decided in courts worldwide. Several unresolved questions affect what you can legally do:

**Training data rights:** Artists are suing AI companies for training on copyrighted work without permission. If courts rule that training on copyrighted material is infringement, it could reshape the entire AI industry.

**Output ownership:** The US Copyright Office has ruled that purely AI-generated images cannot be copyrighted. If you use AI to generate an image, you may not have exclusive rights to it. Courts are still working out where the line falls when humans and AI collaborate.

**Commercial use:** Using AI-generated content commercially (selling it, using it in ads, publishing it) carries legal risks that vary by jurisdiction and are changing rapidly.

The safest approach: treat AI as a collaborator, not a generator. The more human origination you bring to the work, the stronger your legal standing.

These aren't abstract debates. They determine whether the artists, writers, and creators whose work trained the AI will be compensated. Whether AI-generated content carries the same legal weight as human-created content. Whether the economic benefits of AI flow broadly or concentrate further.

You don't have to have all the answers. But you should have the questions.

**Check your understanding.** AI development is concentrated in a few companies, countries, and cities. Why should this matter to someone who just uses AI tools?

- A. It does not matter much — a user only needs the tool to work well
- B. A small group's choices set what AI does well and whose needs it serves
- C. Concentration keeps prices high, since a few firms face no competition
- D. Concentration will fade on its own as models get cheaper to train

## Live Demo

**Step 1:** Think of a topic you care about — something you have a genuine opinion on. Not something you were assigned. Something you've actually thought about.

**Step 2:** Write 3-4 sentences about it. In your own words. Don't use AI. Just your thoughts, your voice, however rough.

**Step 3:** Now ask the AI:

```prompt
Write 3-4 sentences about [same topic].
```

Compare the two. Which one has a point of view? Which one sounds like it was written by someone with a stake in the answer?

**Step 4:** Now try the collaboration:

```prompt
Here's what I think about [topic]: [paste your sentences]. Help me say this more clearly, but keep my voice and point of view. Don't smooth it out into generic AI voice.
```

Compare this to both previous versions. This is what AI collaboration looks like when you bring something to the table.

**Step 5:** Ask the AI:

```prompt
Who built you? What companies funded your development? What data were you trained on? What are the limitations you know about?
```

Read the answers critically. Notice what it says and what it doesn't.

**AI writing alone:** Fluent, well-structured, technically competent. Reads like a knowledgeable summary from someone who's read a lot but hasn't experienced anything. No personal stake. No specific perspective. Could have been written by anyone about anything. Production without origination.

**You + AI collaborating:** Your original point of view, clarified and polished by AI. The voice is yours. The perspective is specific. The insight comes from your actual experience. AI made it clearer and more polished, but the core idea — the thing that makes it worth reading — came from you. Origination amplified by production.

## Why This Matters

AI is not going away. It will get better, cheaper, and more integrated into everything. The question isn't whether you'll use it — it's what kind of person you'll be when you do.

Will you be someone who outsources their thinking to a machine? Or someone who uses the machine to amplify thinking you've already done?

Will you accept AI-generated answers at face value? Or bring the evaluation skills, the data awareness, and the context engineering you've learned to every interaction?

Will you understand who built these systems, whose interests they serve, and what questions aren't being asked? Or will you just be a user?

This program was designed to make sure you're not just a user. You understand how AI works. You can communicate with it effectively. You can evaluate its output. You can build with it. You know your tools. And now you know the biggest truth of all: the thing AI can't replace is a person who has something to say.

## The Challenge — The Capstone

**Challenge: Build and Present** (multi-session)

This is the capstone of the entire program. You will build something real and present it.

- [ ] **Build a project** using AI as a collaborator. It must be functional and shareable.
- [ ] **Solve a real problem** — The project should solve a real problem, express a real idea, or serve a real purpose.
- [ ] **Write a brief** (1 page or less) explaining: what you built, what AI tools you used, what you contributed that AI couldn't, and one thing you learned about AI from the process.
- [ ] **Present it** to the group. Show the project, walk through your process, and answer questions.

Ideas: An app, a website, an interactive tool, a research project, a creative work, an automated workflow, an educational resource, a business prototype — anything that demonstrates what you've learned and represents something you actually care about.

**Success criteria:** You have a working project, can explain your process, can articulate what AI did and what you did, and can answer the question: "What's yours in this?"

## Key Takeaways

1. AI can produce, but it can't originate. Your perspective, voice, and values are what make AI valuable — not the other way around.
2. The top creative humans still far exceed AI. The differentiator is having something to say, not producing fluent output.
3. AI development is concentrated: a few companies, a few countries, a few sources of capital and hardware. This has real implications for power and access.
4. The question "who built this, whose values does it reflect, and who benefits?" is one you should ask about every AI system you use.
5. This program ends with a thing you made, not a test you passed.

## The Rabbit Hole

**Type:** Book
**Title:** The Coming Wave — Mustafa Suleyman
**Description:** The most important book on AI governance written by someone who's actually built frontier AI systems. It asks the hardest question: if we can't stop this wave of technology, how do we maintain any control over it? Endorsed by Daniel Kahneman and Bill Gates.

## References

| Type | Title | URL | Description |
|------|-------|-----|-------------|
| Study | "Researchers Tested AI Against 100,000 Humans on Creativity" (2026) | https://sciencedaily.com/releases/2026/01/260125083356.htm | The landmark Montréal study comparing AI and human creativity |
| Video | Mustafa Suleyman, "AI Is Turning into Something Totally New" TED Talk (22 min) | https://ted.com/talks/mustafa_suleyman_ai_is_turning_into_something_totally_new | The AI pioneer on what's coming next |
| Video | Mustafa Suleyman, "What Is an AI Anyway?" TED Talk | https://ted.com/talks/mustafa_suleyman_what_is_an_ai_anyway | Reframing how we think about AI as a category |
| Report | Stanford HAI AI Index 2025 — Governance & Policy | https://hai.stanford.edu/news/ai-index-2025-state-of-ai-in-10-charts | Data on AI investment concentration and global governance |
| Legal | NYU JIPEL, "Andersen v. Stability AI" analysis | https://jipel.law.nyu.edu | Legal analysis of the landmark artist lawsuit against AI companies |
| Legal | US Supreme Court declines AI copyright case (Thaler v. Perlmutter) | https://futurism.com/artificial-intelligence/supreme-court-blow-ai-artists-copyright | The ruling reinforcing that copyright requires human authorship |
| Tool | GPTZero (AI content detection) | https://gptzero.me | Detect whether text was written by AI or a human |
| Tool | Glaze (artist style protection) | https://glaze.cs.uchicago.edu | Tool that protects artists' styles from being replicated by AI |
| Book | Mustafa Suleyman, *The Coming Wave* (2023) | — | Essential reading on AI governance by a frontier AI builder |
| Book | Kai-Fu Lee & Chen Qiufan, *AI 2041* (2021) | — | Fiction and analysis exploring AI's impact over the next two decades |

## Glossary

- **divergent thinking** — A type of creative thinking where you come up with as many different ideas as possible, like brainstorming ten uses for a paperclip. It's about quantity and variety, not finding one "right" answer.
- **origination** — Coming up with something genuinely new based on your own life, perspective, or a unique insight. AI can remix existing patterns, but only humans can truly originate.
- **AI Act** — The EU's big AI law that sorts AI systems by how risky they are and sets rules for each level. It became law in August 2024, but its rules switch on in stages: bans on the riskiest uses started in February 2025, rules for general-purpose AI models in August 2025, and transparency rules — like labelling AI-generated content — in August 2026. The rules for high-risk systems were pushed back to December 2027.
- **copyright** — Legal protection that gives creators ownership of their original work. AI makes this tricky because models train on people's work without permission, and courts are still deciding if AI-generated stuff can be copyrighted.