#2: Will Generative AI Kill Creativity?

People worry that AI will be the death knell of creativity. I think AI will bolster creativity by bringing in new creators into the market and by expanding the boundary of creative experimentation.

This is a newsletter on AI, entrepreneurship, creativity, and mindfulness. Season 1 breaks down Generative AI and its impact on creative jobs and industries. Post #1 is here.

---

Hollywood writers are on strike. Among their demands is that AI should not used to write or rewrite literary material. There is also the backlash against Marvel’s Secret Invasion using AI’s help to generate the opening credits. There are a range of legitimate concerns around AI’s use in creative fields. I’ll discuss some of them in upcoming posts but this one will focus on the concern that AI use will kill creativity.

The fear is that, with its cost advantages, AI generated content will dominate the marketplace in the same way that mass-manufactured products have replaced handcrafted ones. One only needs to look at the dull uniformity and standardization of modern housing architecture versus the uniqueness of older homes to understand the concern. While I agree that cost advantages of AI will make it pervasive, I do not share the concern that creativity will be strangled. Instead, creativity will thrive in news ways. Before I explain why, let me set some context on how generative AI is impacting creative work.

Generative AI and Productivity

In an MIT study, 444 professionals from various fields completed two writing tasks, including writing press releases, reports, analysis plans, and emails, each taking about 20-30 minutes. For the first task, they wrote without AI assistance, while for the second task, half had access to ChatGPT while the others did not. Researchers assessed writing quality and completion time.

Results showed that the ChatGPT group completed the second task 37% faster (10 minutes less) than the control group, with higher quality evaluations across different measures like writing quality, content, and originality. ChatGPT reduced inequality in writing quality, as the correlation between first and second task performance was weaker in the treatment group, indicating improved skills for those who initially struggled.

Additionally, ChatGPT helped to reduce inequality in writing quality. In the control group, those who performed well on the first task tended to perform well on the second task. However, in the treatment group, this correlation was much weaker, meaning that ChatGPT was helping those who struggled on the first task improve their writing skills. Treatment group reported higher job satisfaction, expressed more enthusiasm about AI's potential for productivity improvement, and had reduced concerns about job replacement.

Now what was interesting is that the scores improved most for people who had poor writing skills initially. So the ones to benefit most were low-skilled workers; high skill workers showed relatively modest increase in productivity.

In another study by Microsoft researchers, GitHub Copilot, an AI tool providing code suggestions, was examined for its impact on developer productivity. 95 Participants were assigned to either a treated group with access to Copilot or a control group without it. They were asked to write an HTTP server in JavaScript. Results showed that Copilot users completed the task 55.8% faster than the control group. Once again, less-experienced developers or those who had a heavy coding load, benefited the most.

These findings distinguish Generative AI from traditional automation which have been the hardest for low-skill workers. Instead, generative AI seems to help low-skill workers “punch above their weight.” The impact is not just a general increase in productivity but also a narrowing of the productivity gap between low-skilled workers and high-skilled ones. This could fundamentally alter the creative marketplace as I discuss below.

Generated using Lexica.art

Expansion of Creator Economy

One of the biggest implications of generative AI for creators will be the expansion of the creator economy. With AI tools making content creation more accessible than ever before, we're likely to see a massive increase in the number of creators. AI-based onboarding and education will also help further lower barriers, meaning that almost anyone with a computer and internet access can become a creator.

To get an idea of just how transformative this could be, consider the music industry. During the Internet boom, digital tools made music creation more accessible than ever before, leading to a nearly 10x increase in the member base of performance-rights organization ASCAP, which represents 900,000 songwriters, composers and music publishers. It's entirely possible that generative AI will drive another major increase in music publishers in the coming years.

The democratization will not just feature lowering of barriers and entry of new creators but will also feature a shift in industry power dynamics (more on that shift in my next post).

Will AI Kill Creativity?

I believe creativity will thrive in news ways for at least three reasons.

First, as I just mentioned, generative AI brings in more and new kinds of creators into the market. We will therefore see a greater diversity of creative output.

Second, AI will often serve as a tool for creators and significantly enhance their productivity. Increased productivity will make it cheaper and faster for artists to experiment. Where it might have taken years to run creative experiments in the past (e.g. a writer exploring multiple alternative endings for a screenplay or exploring different conflicts for a central character), it can now be done rapidly. The time that AI will free up from activities like rough drafting can now be used towards these kinds of experiments. Where artists might have given up on experimentation and wrapped up their creative process due to fatigue, deadlines, etc, they will now be able to continue exploration. Additional experimentation and explorations will help unleash greater creativity.

Finally, even in instances where content is created primarily or entirely by AI, it is a misnomer to believe that AI is merely a “plagiarizing machine” that will primarily recreate styles from the past. Modern generative AI leaves enough room for AI to randomize and experiment. For example, AI might be better able to mash up different creative styles to generate completely new art styles that may be hitherto unexplored; after all, a lot of human art today still involves combining creative styles in unexpected ways. Further, Reinforcement learning – a class of machine learning algorithms that learns by doing as opposed to learning from historical data – has experimentation built into the learning methodology. In short, while I do not doubt that AI will experiment differently from humans, it will nonetheless be able to experiment and surprise us with new and very creative outputs.

The invention of photography was met with concerns from many artists. But in the end, it not only created a new form of art (photography itself) but also became a tool for painters and “released painting from the need to be realistic.” Human artists evolved from only representing reality to also representing emotions (hence, the rise of impressionism). Art will similarly evolve with generative AI.

---

So, how will AI impact wages for creative work or the industry power structure itself? I’ll leave that for my next post. I'll also save my non-technical tutorial on generative AI for a future post.