Creative Intelligence II
My experience with AI-assisted writing
AI-assisted creativity tools are all the rage. Or perhaps they’re just my latest obsession. If you’re new to the concept, allow me to summarize: Machines are turning into collaborators for creative activities.
Up until this point, our creative tools have been entirely command-driven: Copy this. Paste that. Apply a filter. Our favorite platforms (Figma, Canva, Adobe Premiere) have boosted the creative process by adding new features, and expanding the range of commands a human can perform. Under this paradigm, any use of machine learning has merely been… a ‘supporting character’: AI for noise reduction, AI for rotoscoping, etc. But these are still just commands. A human was always in the driver’s seat. In this paradigm, the idea of the machine as a creative agent is non-existent. But, suddenly, it’s everywhere.
I’ve been down a rabbit hole the last few weeks playing with AI-enabled creativity tools, mostly to tools to recalibrate my understanding of this new frontier. For writing I’ve tried Copysmith, Copy.ai, and Lex. For image generation I’ve used OpenAI’s DALLE-2, Midjourney, and Stable Diffusion. And I have to say… I’m damn impressed.
I not only see the immediate applicability of such tools, but also the opportunity for their dramatic improvement. Even if we were to freeze progress on the underlying AI, there is plenty of headroom for value creation on the user interface alone. But of course both are happening simultaneously. Leading to a fascinating landscape of companies trying to seize marketshare and mindshare across all layers of the stack. How can we understand, perhaps, predict, the epic capitalist drama that’s about to unfold? That’s what I’m here to find out.
Prediction & Creativity
When machine learning first appeared on the scene, it was obvious that society collectively struggled to grasp the implications of this new technology. And I don’t just say this in hindsight either. In the moment, we were painfully aware that we had certain blindspots. We simply didn’t have the mental models to conceptualize what impact of ML. Today, now firmly in the midst of the AI-revolution, one framework has transcended them all: AI introduces low-cost prediction.
The key insight here is that ‘prediction’ is an economic good, an input as tangible as steel or energy. For too-long our use of prediction was supply-limited by the availability of human workers. Only mathematicians could reliably pattern-match large data stores, and only specialists could deliver services like object detection and face recognition. Now that computers can perform these tasks, the supply of these services has exploded, and the cost of such predictions is near-zero. Today, every website and app is ‘predictive’, empowered by ML to be more attuned to the needs of each user. We’ve now learned that lower cost creates ubiquity, allowing for certain services in the economy to be consumed at orders of magnitude greater scale.
We can extend this framework by saying that the use of generative ML is introducing low-cost creativity. We briefly touched on this concept in our last article. Now let’s go deeper.
I don’t have to convince you that creativity has economic value. Creativity is the bedrock of several multi-billion dollar industries: fashion, entertainment, and advertising. The top 10 companies in fashion and entertainment alone have $920B and $807B in market cap respectively. Now what will happen to these industries as the creativity approaches ‘free’?
Our initial response is to assume companies will cut costs and replace their writer’s room with a server rack. But I firmly believe this is not the path we’re on. Instead of imagining a society without creatives, we should picture one with millions more. Cheap access to generative AI models will allow everyone in their dorm room - hell, even kids in elementary school - to channel world-class creativity. Software always has more of a democratizing influence than a centralizing one. Remember, lower costs drives increased consumption. The future will be one where the economy has 100, 1000, maybe even a million times more ‘creative resources’ present.
You may wonder, then, what will happen to existing creatives? How can they be expected to live in a world of such cutthroat competition. To answer this question, we need to understand more about the evolving relationship between humans and creative machines.
In The Loop
While creatives will always live among us, our definition of the creative class will change. Today, a designer/architect/digital artist sits down at their computer, and clicks around until they make something worthy of their craft. With AI, you treat the computer as a sort of conversational partner.
I used the AI-writing tool Lex to produce a rough draft of this article. When I was tired, stuck or even just plain curious, I would prompt the AI to make a suggestion. Some of them were good enough to make it into the draft! Compared to writing past articles, I can honestly say I really enjoyed the experience. There was something simple and lighthearted about the back and forth.
As I neared publishing, however, I subjected my draft to a standard last-mile of refinement. Reading this article now, Lex’s voice is nowhere to be seen. I believe the key advantage is that Lex allowed me to converge on an end-to-end draft much faster than I otherwise would. This is well-captured in the below chart. Generative writing models gave leverage to my moment of inspiration, allowing me to produce more content in the neighbourhood of my thesis.
Is this the same article I would have written had I not used AI? In the important ways, yes. In the nitty-gritty details, probably not. There is undoubtably a ‘butterfly effect’ here, where the more you rely on your AI partner the more your article inches away from your original intent. But you, my reader, can trust that I’ll stay true the overarching narrative in my head and try to deliver a few pre-determined insightful points along the way.
What machines are really capable of is raw production, and this is well-suited to some parts of the writing process. For any article I publish, I typically leave 3-5x as much content the cutting room floor. The majority of my time is spent iterating over ideas and paragraph structures. With AI, I can get exposure to combinations of words, and alternate ways of communicating at a lower personal cost. Instead of racking my brain for alternatives, I can simply generate them and filter for my preferred configuration.
Now consider this same principle applied to other creative areas. AI has the ability to reduce the development cycle of any creative process. A fashion designer might spend dozens of hours producing potential outfits in a sketchbook, and even more days physically prototyping with different cuts and materials.
Instead of spending dozens of hours sketching outfits in a scrapbook, AI could let a designer generate and edit clothing in 2D or 3D space. It’s easy to picture our designer producing 10x more clothing concepts than their peers. But at the end of it all, what defines a successful product is the keen eye of the creator. It’s their ability to manifest something that is indescribably, culturally resonant and cool that matters.
This is why I believe we’ll no longer define creative work by the labor of creation. If a hundred more hours results in a better outcome, you’re creating an artisan good. But if you’re making a routine cover photo/blog post that could be generated by AI in 1/10th the time, I would call that economically unproductive. When you remove effort from the equation, what is steadfastly valuable about the creative class is their unique sense of style or taste.
A Human Touch
The puritans among us may still think these tools are blasphemous. They will insist that there is something uniquely human about creativity. And I agree with them.
Narrowing our scope to just writing: In today’s age of ubiquitous content marketing, we mainly experience writing as the soulless way companies’ generate leads in competition for Google keywords. Or as the nonsense storytelling before an online recipe so a site can show us more ads. This is the dark side of ubiquity. It makes us forget how this advance could be used for good. Real writing has a sacred purpose: to contribute to the Great Conversation.
In my new favourite Wikipedia page, the Great Conversation is described as: “the ongoing process of writers and thinkers referencing, building on, and refining the work of their predecessors.”
This is something machines have yet to accomplish convincingly. In my experience across all these tools, an AI never gave me an idea that wasn’t originally intended to be in the piece. Models are trained on past data, which means novel idea generation is something that the current paradigm of AI is not capable of beyond pure randomness. (But I am open to the idea that future AI’s will be able to re-conceptualize ideas!)
For now, human creators will fulfill the human need for authentic content. Like the oral storytellers of old, they will curate fables and narratives from the mass repository of human experience. And good content - real content - will rise to the top because of our natural ability to determine good from bad taste.
I am optimistic about what AI tools will do for humanity. Society ebbs and flows based on our ability to share and collectively develop ideas. Technology causes the world to accelerate at an ever-increasing pace, and quite simply - we are struggling to keep up. But we must find a way, and that will require finding better ways to externalize our opinions.
If there was an argument for AI-assisted writing, it’s this: generative AI models promise to lower the barrier of entry to writing and publishing, providing the intellectual leverage that let us all join the Great Conversation. If we let our discussion be dominated by only those voices with entrenched power and influence, then we will really be sacrificing our creativity.
And yes, AI helped me write this outro.
If you enjoyed this article consider subscribing for weekly content on the technology that will shape our future!
If you really enjoyed this article consider sharing it with an intellectually curious friend or family member!
Let me know what you thought about this piece on Twitter.
With gratitude, ✌️