Human cognitive labor will survive
I am not afraid of an army of lions led by a sheep; I am afraid of an army of sheep led by a lion.
Alexander the Great (supposedly)
Let’s say LLMs become 10x as smart as the brightest humans: alien-level intelligence for every domain, replacing legal counsel, engineers, etc.
At first glance, with those tools, it seems like a single person could run entire companies. “Human cognitive labor becomes worthless. The mass of humanity is cast aside.”
Sounds grim. But let’s think. The company leader and LLMs need to be aligned. What does it mean to be aligned? It means understanding, and agreeing on:
- why we’re doing something
- what we’re doing
- how we’re doing it
- when we’re doing it
- who’s doing what
“Who’s doing what” is easy to answer: the LLMs do all the implementation work, while the leader tells them what to do. We can hand-wave “why we’re doing it” to be “maximize capital,” and “when we’re doing it” to be “immediately, 24/7.”
The “what we’re doing” question opens the first can of worms. The leader is in charge of what to do. Thus, human judgment is inescapable: the human decides what to work on.
“But they could ask the LLM what to do!” In that case, the leader still exercises human judgment when picking between the LLM suggestions.
“But they could just fully delegate to the LLM!” Fully delegate the LLM to do … what? To do something the human prompted. The original delegation command, no matter how vague or meta, still falls on the leader to create. There is no escaping this first domino.
We find that human judgment remains essential, and that its value is independent of how strong LLMs become.
Let’s look at the next question: “how are we doing this?”
If the leader fails to understand how the LLM is accomplishing the goal, they’ve become misaligned.
This misalignment dooms the endeavor. The human increasingly loses touch with what’s happening, and the human ironically becomes the one who hallucinates about the actual state of affairs.
To align with human ends, the requirement is human understanding of the work, which requires human expertise.
By this line of argument, we’ve found that we still need both human judgment and human expertise, independent of AI strength.
If this holds true, human cognitive labor still holds value, and thus its market will survive even godlike LLMs.