Transcript: Why cheaper AI may not lower living costs
An AI-generated audio overview (Google Gemini), transcribed by machine on Blaine's Mac (Whisper). The audio was checked against primary sources; what it gets wrong is listed at https://blainomd.com/learn#audio-why-ai-makes-life-expensive

Corporate profits are hitting record highs at over 11% of GDP.
But the value of human labor is plummeting faster than at any point since the 1940s.
So welcome to the deep dive.
Today, our mission for you is to unpack this massive phase transition that's hiding in plain sight.
We are calling it the Great Decoupling.
Yeah, it's a huge shift.
We're looking at a stack of economic and technological analyses, and they all show the cost of intelligence itself is collapsing.
It's dropping by about 100 times per year.
100 times.
Right, per year.
So economic productivity is being completely severed from human labor.
Basically, we are taking the entire knowledge economy and turning it into a digital commodity with near euro marginal costs.
I mean, it really reminds me of the transition from horse-drawn carriages to the combustion engine.
Like when the car came along, we didn't train the horses to do a new job.
No, they were just retired.
Exactly. Retired from the economic equation entirely.
And AI is starting to do that to complex reasoning and creative output.
But here is the massive contradiction staring us in the face.
A cost of living.
Right. If cognitive labor is basically free, why isn't my grocery bill dropping?
You know, why isn't everything getting cheaper for you and me?
Well, because we're seeing the deflation of thought colliding directly with the hyperinflation of physical assets.
On the digital side, AI has erased what economists call price stickiness.
Price stickiness, right.
Yeah. So before, a store might keep prices steady because, you know, updating physical tags was a hassle.
Plus, competitors might undercut them.
But now, algorithmic pricing bots instantly monitor competitor pricing.
Do they just raise prices together?
Instantly. In milliseconds.
That lag we used to rely on for cheaper goods is totally gone.
Wow. So the digital world gets flawlessly optimized. But, I mean, you can't optimize concrete into existence.
AI can design the perfect skyscraper in 10 seconds, but it still takes steel, land, and energy to actually build the thing.
Exactly. And the AI revolution itself demands massive physical infrastructure.
We're talking about information power plants.
Like sprawling data centers.
Data centers, silicon foundries, massively upgraded energy grids.
That creates a staggering demand for capital, which keeps interest rates incredibly high, even as the cost of a digital service plummets to zero.
Hold on, though. Software usually drives overhead costs down across the board.
If AI is making systems so hyper-efficient, why are highly necessary life-or-death services like health care facing skyrocketing costs?
It's a great question.
It's like, why can't a hospital just replace its massive administrative layers with AI and pass the savings on to you?
Because health care faces what we call a compounding reliability trap.
Imagine an AI agent with a 95% success rate per step.
Sounds fantastic for writing an email.
Right. But over a 20-step clinical workflow, that end-to-end success rate drops to a dismal 36%.
You still absolutely need human validation gates at every critical turn to ensure patients don't get hurt.
Ah, I see. So we still absolutely need humans in the loop.
But those humans now expect higher salaries because, well, they could just go work for a massive tech company instead.
That is Bommel's cost disease in action.
Health care requires human time and empathy.
To stop doctors and nurses from fleeing to hyper-lucrative tech sectors, hospitals have to dramatically raise wages.
But they don't have the software margins.
Exactly. They raise those wages without matching the massive productivity gains of a software company.
It creates this totally unsustainable wage wedge.
Which leads us to a really uncomfortable structural reclassification of human utility.
Economically speaking, we're transitioning from the factory workers generating wealth to the expensive heads consuming it.
That is a stark way to put it. But yeah, historically, humans were a nation's primary asset, the engine of wealth.
Right.
But in this new intelligence century, wealth is generated by compute, not human brains.
Combine that with fertility rates in tech hubs dropping to between 1.3 and 1.8.
People just aren't having kids.
Right. So governments face a shrinking tax base and an aging population.
We transition from being an economic engine to a fiscal cost center.
And those cost centers need a health care system that is currently, as you said, pricing itself out of existence.
The immense fiscal pressure is pushing traditional health care toward a balloon pop insolvency.
We'll likely see it in the late 2020s.
The monolithic hospital system simply cannot sustain the wage wedge.
So what happens when it pops?
It will rupture and force a shift toward decentralized, edge-based care.
Edge-based care.
Yeah. Think along the lines of a seven-hour nanopore genome sequencing done right at your kitchen table instead of in a billion-dollar centralized facility.
Wow. Okay, so we've traced this decoupling from the deflation of knowledge work straight through to the inevitable balloon pop of our health care systems.
We really are stepping into a creator era where wealth belongs entirely to those who own the grid and the compute.
And if our value is no longer in our labor, everything hinges on how the AI is built.
Will future AI possess thick alignment?
Meaning a deep contextual understanding of human flourishing and empathy.
Exactly. Or will we just get thin alignment, where the AI operates perfectly, but views humanity as nothing more than an inefficient biological cost center consuming its resources?
That's the billion-dollar question for you to ponder.
The machine doesn't care who pushes the buttons as long as it runs the numbers.
Will the ultimate intelligence see us as the purpose of the equation, or just a rounding error to be zeroed out?
We'll leave you with that thought until next time.