on the nature of expertise
and why it may be all you need
Sitting with a friend over dinner before the 2025 holidays, we were discussing the nature of economies and arrived at a topic that has since lived in my head rent-free. What distinguishes expertise from knowledge? We analyzed expertise in a global and historical context, stemming from Industrial-era British textile production to globalization-era Dutch lithography1. Conversely, the post-war bloom of American universities built a population armed with knowledge, and today’s Gaokao and JEE exam systems set up China and India respectively as human knowledge superpowers.
To me, though, the question became more fundamental and more personal, and I set out to answer it. Here’s my take.
As I see it, knowledge is simply information that is agreed upon. Through varied doses of both scientific method and consensus, we build the web of knowledge. Expertise, or its cousin wisdom, can be defined as using lived experience to apply knowledge situationally.
The distinction between knowledge and expertise is core to the modern human struggle in the AI age. Public opinion ranges widely on how AI progress will develop in a way that humans cannot compete, with direct implications on how humans should be educated and what skills humans should acquire.
As AI crushes legal exams, derives new mathematics, and co-invents new biological research, it is clear that today’s AI can remember, apply, and create new knowledge. However, false confidence, lying, and consistent lack of context are a far cry from the establishment of expertise.2
Humans have traditionally been valued for both their knowledge in subjects as well as their expertise in their fields. Evaluations of both humans and AI can correctly assess what is known at a given point in time. Through it all we rely on human beings to assess true capability with proper judgement (or “taste” as it is often referred to within the AI industry).
Judgement, or taste, or wisdom or expertise, is extremely hard to quantify. Ask anyone in any discipline: assessing capability is full of flaws. We anchor on mostly two archetypes: a single prove-it moment (one exam or one interview) or counting stats (hours flown, years of experience, patients seen, points scored). This exact quality of expertise is what keeps it safely in the hands of humanity.
Until AI is demonstrably and measurably better than humans at a given task, humans will not relinquish their control. Because expertise evades measurement, the burden of proof rests on AI to show it has more expertise than some “certified” humans.
It is now logical to me that domains of knowledge will be firmly dominated by AI, while domains of expertise will be strongly held by human beings.
On the nature of expertise, I have come to the following conclusions:
Internalize that human dominion over knowledge is ending. If you are a curious human being, sit with the discomfort this may cause.
While knowledge is not useless, it is prudent to acquire knowledge primarily in the interest of developing expertise.
The ability to gain expertise, and the rate at which you can acquire it, is critical to your ability to both survive and thrive.
Expertise evades measurement. For both artificial and human beings, there is no reliable method by which to measure expertise.
As far as we know, expertise is acquired through lived experience and social proof. Therefore, as of today, humans are uniquely capable of expertise.
In other words, Mom told me to make myself useful.
I love San Francisco.
Today’s AI lacks the capabilities to learn continuously, improve itself autonomously, and coordinate amongst itself. It is a matter of debate if a combination of any and all of these traits will lead to a development of expertise.