If AI Can Make Anything, Taste Matters More Than Ever

by: Austin Spires

My wife and I will welcome our second child into the world in approximately a week. This post initially started as a thought exercise on what "career advice" I would give my children in approximately 13-15 years.

On its face, that’s an absurd goal. No one has been able to reliably or accurately predict the world’s macroeconomic circumstances with detailed accuracy. Or at least in a way that can provide a concrete playbook on life. Though, as my CPA father will point out, the world has needed accountants for hundreds of years and will likely continue to need accountants for hundreds more. Add math for good measure, and you’re good — double major in accounting and mathematics. If you can survive that gauntlet, the world is your oyster. If you really want to seal the deal, take that pristine double major and become a tax lawyer.

However, the real catalyst for my thinking, beyond the massive responsibility of shaping two children into becoming their best selves, is the current AI Revolution that we’ve found ourselves in and the constant chatter of "AI is killing <$profession$>" with editors and controversy-farmers replacing the profession with whatever might generate the most chaotic attention.

What is this "revolution"?

If you can’t tell my opinion on these statements, I don’t think AI will decimate millions of jobs in the next 3 years or eliminate high-quality professions overnight. Every time there’s a significant breakthrough in technology accessibility or cost decrease that promises leaps in productivity — whether it’s the printing press, the word processor, the industrial revolution, agricultural innovation, the internet, or the internal combustion engine — hype cycles have simultaneously preached utopian futures with no work or hardship powered by the technology and dystopian societies where everyone suffers at the wrath of the powerful few who amass global control of the technology. Jevon’s paradox and Malthusianism are good Wikipedia reads to learn how long these arguments have been going on. Neither extreme comes to pass. Access and affordability improvements make things better and worse in different ways.

This is what our current AI Revolution really is — a massive breakthrough in making these technologies accessible and affordable for the first time in history. LLM models, image generation, machine learning, information retrieval, Markov processes, and similar techniques have existed in lab and research settings for decades. They’ve also been highly fascinating for decades, to explicitly declare that I love technology innovation broadly and have excitedly followed AI breakthroughs for years. Those breakthroughs have never been accessible through a bulletproof, freemium, intuitive, and responsive chat and API interface until now. And that interface and accessibility breakthrough deserves the hype it receives.

Some of the concerns about AI’s economic destructiveness are valid. In particular, I think about Jessica Hische’s exploration of the potential impact of AI on her profession, as well as the current trend of vibe coding to build applications whole-cloth from prompts. What recent ChatGPT and AI breakthroughs will have on the next generation of lettering artists, programmers, writers, video producers, lawyers, teachers, and every other profession is unclear.

But here we are: robots can do more than we’ve ever anticipated, and their capabilities are rapidly expanding. What now?

What do humans do in an AI world?

If we’re moving to a futuristic interface with computers where we can send inputs using everyday conversation and receive outputs extremely quickly, it will become a skill to ask the right questions, issue the proper commands, and (most importantly of all) evaluate the output. Breaking that down, it requires:

1. Having an impactful, unique, or meaningful idea.
2. Articulating that idea in a clear, understandable way.
3. Reviewing the results in both a quick and accurate manner.
4. Iterating on that idea to refine the outcomes.
5. If necessary, wholesale scrapping of the strategy and restarting.

These skills are difficult to achieve in practice and could take years of experience to perfect. But they each require something that a computer will never have by design: taste. A computer will never have compelling taste in the way a human can. Developing taste is complex and challenging, but it’s the only way to rise above the threat of a robotic or automated tool taking one’s profession.

What is taste?

A trusted mentor of mine constantly uses the phrase "know what good looks like; describe what it is, and how to get there" as a leadership principle. This is the most practical definition of taste in a professional setting, and it’s been one that I can rely on constantly. It works particularly well for leading teams of people, as the mindset empowers groups of talented people to elevate their collective work to break ground into something new. Of course, the point is not the destination, but building the culture within the team to strive for better as a sum of their parts. This is a daily practice and requires constant attention, reinforcement, and self-discipline to stay on track.

A different definition, originally from creative crafts and the arts, is "knowing what’s going to be exciting one year in the future, and pulling that forward to today." This is great for knowing what will resonate with audiences, customers, readers, users, or "lots of people." It’s more destination-oriented than process-oriented, but that doesn’t mean it’s less critical. The brutally difficult part of this rule is the one-year window instead of the distant future. It requires a deep understanding of people: what they like, what they don’t like, what makes them fearful, what makes them excited, and how you can tap into that psychology immediately. It also requires knowing the inner workings of your medium well enough to know what can be created in the near term. Science fiction has no shortage of extraordinary visions for the future: instantaneous travel, unlimited energy, and androids with rich personalities. Academia has similarly ambitious visions as well. This vision is necessary for the advancement of society, but connecting that distant future into something populations can quickly adopt within one to two years is a missing piece in turning these hypotheticals into reality. It’s extraordinarily difficult to connect the future to the present. Some get lucky once. Fewer can do it more than once. Strive to be able to do it more than once.

This is a good segue into a third definition of taste that's even more future-oriented. Jim Simons, the founder of Renaissance Technologies, the original and most successful quant hedge fund, said in an interview with the New Yorker in 2017 that "taste in science is very important…to distinguish what’s a good problem and what’s a problem that no one’s going to care about the answer to anyway—that’s taste. And I think I have good taste." Business analysts might call this "picking the right market." Artists might call this "chasing a willing audience." When I was an athlete, coaches called this "working smart and hard, not just hard." In short, you need to know where to focus your energy to benefit yourself and the population you’re working with. Simply toiling is not enough; you must discern where your toil can have the most leverage.

These three definitions fit very well as a funnel from most strategic to highly tactical. Taste requires knowing the broad, audacious problem space to orient one's career and craft; the near-term milestones to bring the future into reality in an accessible way that allows you to pursue the long-term goal incrementally; and the mindset and standards to move toward the milestones through the high and low moments of life on a daily basis. If these three can exist in harmony, the presence of AI is just another vehicle to execute on the strategy, not a blocker.

However, a fourth component is also essential — as if mastering the first three isn’t tough enough. You have to quickly tell the difference between art and imitation. Framed another way: in the same way that taste is knowing what good looks like, it’s also knowing what bullshit looks like and calling out that bullshit right away. Simplified further, you have to "know shit from Shinola". "Slop" has become a common term for derivative AI-generated work, though every creative trend or innovation breakthrough has had its fair share of slop. Also, not all AI-powered work created today is slop. Like all other paradigm shifts, extraordinary work transcends the medium and is immediately relatable. The fluency required to navigate new trends while quickly discerning slop from quality separates the people who master their tools from those dictated by them.

So, to my kids 15 years into the future, I have no idea what’s happening now in your worlds. I have no idea what the cool innovations are, if AI is still around, what the hottest college major is, or even if college as I know it still exists following the AI revolution. What I do know is: develop taste and discernment. No AI will ever be able to replace that skillset.