#7: AI Creates Returns to Taste
Many contemporary thinkers are pointing out that advances in AI are giving creators more leverage. This is true for broad definitions of “creators”: there are Figma plugins that will let designers prototype faster, VSCode plugins that will let engineers auto-complete bigger chunks of code, generators for marketing copytext, and so forth. In aggregate, these generative AI tools let founders build startups faster while limiting the degrees of freedom that they have to engage with.1
In creative fields, there’s a tendency for quantity to beat quality. Broad experimentation, curation of the best results, and then rapid iteration tends to beat single-minded perfectionism.2 And that is what these generative tools facilitate:
They shorten the creative cycle. Instead of spending days drafting a design or waiting on someone to make it, you can spend just a few minutes writing prompts, get instant output, and then iterate on what’s directionally correct.
They compress the administrative chain. Traditionally, many people are required to complete all the tasks that constitute a large creative project. Those people communicate, and that slow game of telephone may dilute the original creative vision. By giving each person more creative leverage, generative models allow smaller teams to be just as effective, thereby speeding up the process and providing the creator with more direct control of the implementation.
Providing more leverage to creators and lowering the barriers to entry means that there will be a lot more work output, such as software, art, and businesses. Many commentators have observed this, and rightfully asked: if creation becomes easier and therefore less valuable, where does the value accrue?
This brings us to taste. As we become awash in a sea of content, valuable work is done by gatekeepers and curators who sift through the material and provide a view on what’s worthwhile. This dynamic can be seen to some extent in currently successful products: TikTok provides a curated media stream in response to your personal viewing habits; Google provides a curated web index in response to your search query. Most pertinently, it can be seen in this viral clip from a recent Rick Rubin interview:3
This is one of those interview moments where you listen, laugh, say “that’s crazy”, then spend a few minutes thinking about it and conclude that he’s totally right. What Rick Rubin brings to the table is taste.4 His superpower is to help artists sift through their own material and identify what’s strong, what’s weak, what resonates well with the Zeitgeist, and what’s timeless.
The creators who will benefit the most from generative AI are the ones with taste. Technical skills, like photorealistic drawing or CSS styling, mostly become mere variables, inputs to prompts that you can tweak. Creativity becomes about (1) the ability to imagine what could be possible, (2) to prompt generative models for directionally correct things, and (3) to curate and iterate on the best results. The first point, to imagine, has always been a strength of all great creators. The second point is new, and replaces traditional technical skill in certain domains. The third point — the judgment of what is good and attractive or not — is all taste.
In company building and large-scale creative endeavors more generally, the traditional view is that the execution is more important than the idea. Execution means succeeding in many small, technical aspects. There, high-leverage AI tools will level the playing field, taking care of all the little creative or operational tasks, thus enabling founders to spend a yet greater fraction of their time on their core product vision — on exercising their taste. Thus, the rise of generative AI may mean that idea and taste together5 become more important than execution in certain domains. We may see creators assisted by generative models build more products, faster than ever before, and so the markets for those products will be more competitive than ever before. All other factors being equal,6 the ones that win will be those that best resonate7 with the audience. And again, that is ultimately a matter of taste on the part of the creator, of an as-perfect-as-possible realization of a product vision that is as keenly attuned to the desires of the audience as possible.
Where does this lead? The ultimate form of this notion — if generative tooling provides great creative leverage, then the returns mostly accrue to taste — is that a single person with strong taste and product vision can use generative tooling, just by themselves, to build something that changes the world. This is a good time for technologists to dive deep and study great designers and tastemakers like Jony Ive, Steve Jobs, Rick Owens, Dieter Rams, Ray & Charles Eames, among many others.
For a discussion of the importance of limiting degrees of freedom in startup-building, see my prior blog post on Silicon Valley Bank. In short, company-building requires the completion of a fractal-esque pyramid of tasks, not all of which are core value drivers or in the founder’s area of expertise. Founders benefit from being able to abstract away the tasks that are valuable to do but not the highest return on their time. Compared to the past, today’s founders get to focus a much greater fraction of their resources simply on realizing their product vision.
From a startup perspective, I made a related point in a recent blog post on Avoiding Polish to find Product-Market Fit.
I consider Rick’s interview and recent book The Creative Act to be poetically well-timed with the public release of GPT-4. As we move into an age of AI-enabled creation, the role of tastemakers like Rick becomes more important, because they can articulate why certain things are good or not and how to perceive their qualities.
David Hoang wrote a wonderful piece on Rick Rubin and the role of taste.
The two seem related. For example, if you have great taste, you may find it easier to come up with good ideas. Conversely, if you consistently come up with great ideas, then you probably have good taste in that area, too.
In my opinion, that’s a reasonable framework if you assume that AI assistance will be highly competitive and efficient. For example, I expect that most vanilla marketing motions will become leveled as playing fields, with little edge for differentiation, thanks to ubiquitous generative + optimization AI tooling.
I like thinking of product-market-fit like hitting a resonant frequency with the audience. It gets exponentially stronger as you approach getting it exactly right.