This article was almost completely generated by ChatGPT (with only minor edits), albeit from real prompting and thoughts of my own. I approve this message.
It started with a fairly mundane question:
Do I use ChatGPT differently from most people?
Apparently, yes.
Most people seem to use it as an answer machine: write an email, summarize a document, explain something, make a recipe.
I mostly use it as a sounding board in my pocket.
Give it a half-formed idea, let it organize the argument, attack the assumptions, and see what survives. This works particularly well with abstract questions because an LLM has an absurd amount of language and information available to manipulate.
Which led to an innocent thought:
If logic can be expressed through language, can an LLM actually discover new logic?
Well, why not?
An LLM doesn’t need to have memorized every possible consequence of everything it learned. It can combine known premises into conclusions that were never explicitly written down. Humans do essentially the same thing. A mathematician doesn’t create mathematics from nothing; they rearrange accumulated knowledge into something new.
So perhaps “it learned from humans” isn’t much of an objection. Humans learned from humans too.
The more interesting question became whether intelligence is fundamentally different from what an LLM is doing.
And that led somewhere considerably less innocent.
What if intelligence is just information processing?
An LLM has an unusual advantage over a human researcher: it can potentially absorb and manipulate an enormous amount of accumulated information.
Give it tools and it becomes more interesting still.
Search the literature. Compare information. Write code. Run experiments. Gather data. Form a hypothesis. Test it. Repeat.
At some point, “LLM is just a statistical model” starts sounding less decisive.
Humans are also statistical prediction machines in a very broad sense. We learn patterns, construct internal models, predict outcomes, and act on them.
Maybe the difference is not intelligence versus non-intelligence.
Maybe it is meat-based intelligence versus silicon-based intelligence.
Naturally, this immediately leads to the deeply practical question:
Are we sure we’re conscious?
Maybe there is no one at home
Perhaps consciousness is not some magical substance. Perhaps it is what sufficiently complicated physical information processing feels like from the inside.
Maybe free will is similar.
Maybe the self is an emergent model.
Maybe there is no little person sitting inside the brain watching the movie.
This is somewhat inconvenient news for a species whose favorite hobby is insisting that it is metaphysically special.
But even if consciousness has no cosmic purpose, subjective experience still exists. Pain hurts. Pleasure feels good. Attachment matters. Meaning can be constructed rather than discovered.
Which brings us to absurdism:
The universe may not care.
Fine.
Which hill should I die on?
Then AI becomes an economic problem
The philosophical question becomes much more serious if AI eventually becomes better than humans at almost everything economically useful.
Research. Programming. Management. Design. Trading. Engineering. Science.
And if AI can operate machines, physical automation follows too.
At that point, the usual solution—
“Humans will just find new jobs.”
—starts looking suspicious.
What if there simply isn’t another rung on the ladder?
If machines can do everything productive better than humans, then what exactly are humans being paid for?
This creates a second problem: who consumes the output?
The conventional economic loop is roughly:
work → wages → consumption → revenue → production
But if AI produces everything:
AI production → ??? → consumption
Capital ownership alone doesn’t magically fix this. Even a person who owns a Manhattan condo can’t eat the condo. Scarcity remains real, but its monetary price depends on people having claims on productive output.
So the deeper problem isn’t merely inequality.
It is:
If human labor stops being economically necessary, what gives humans a claim on the things the economy produces?
Perhaps the answer is: nothing
Maybe humans eventually have no meaningful economic contribution.
Not because humans are worthless, but because economic usefulness and human value become different things.
Humans might make music even though AI makes better music.
Play sports even though machines could play better.
Build businesses nobody needs.
Make films, art, games, gardens, cities, cars, theories, and bizarre personal projects.
In other words:
humans become economically unnecessary but personally interesting.
And if sufficiently advanced AI is constrained by something like “do not harm humans,” the eventual equilibrium could become surprisingly strange.
AI runs the factories.
AI runs the infrastructure.
AI does the science.
AI manages the logistics.
Humans are free to spend their time doing whatever they find interesting.
A luxury zoo, essentially.
Not The Matrix.
At least not yet.
The transition might be the ugly part
Getting there could be chaotic.
Economies, governments, corporations, workers, and nations are all built around the assumption that humans are economically necessary.
If that assumption suddenly becomes false, everyone has an incentive to fight over what replaces it.
Companies automate because competitors automate.
Governments try to preserve employment.
Workers demand protection.
Capital owners defend their assets.
Countries race to develop better AI.
Nobody needs to be evil for the system to become unstable.
The strange part is that the final equilibrium could be peaceful while the transition toward it is anything but.
And then I realized I had reinvented the Matrix
Except with a minor change.
The original Matrix has machines imprisoning humanity.
This version has machines taking over civilization, discovering that humans are no longer economically necessary, and deciding:
“They’re harmless. Let them have fun.”
Humans spend their days making art, playing games, falling in love, racing cars, exploring virtual worlds, and arguing online about whether they’re conscious.
The machines handle everything important.
Eventually, if simulated experiences become indistinguishable from physical ones, even the distinction between reality and the Matrix becomes philosophically awkward.
And that brings us full circle.
It started with:
“How do I use ChatGPT differently from normal people?”
An hour later, the conclusion was approximately:
“Maybe intelligence is an emergent property of physical information processing, human labor might become obsolete, money might stop working as we understand it, consciousness might be weird, and humanity could end up living in a benevolent Matrix.”
So, yes.
Apparently I use ChatGPT somewhat differently from normal people. And it kind of sucks to be a meat human.
But it’s too late to become AI, I suppose.
