January 28, 2025 at 06:26PM
Anyone who is watching AI closely is worried about the future because it will bring massive disruption in the jobs market, particularly if AI unlocks robotic workers, which I believe it will.
Now I’m normally an optimist but I’m having a hard time figuring out how this ends well for most workers. Here is a take from Netscape founder and venture capital investor Marc Andreessen that highlights a particularly positive view.
A similar thinking is shared by OpenAI CEO Sam Altman who talked about a world of abundance or a “post-scarcity” economy where basic needs like food, housing, and healthcare become extremely cheap or nearly free. He has talked about a re-writing of the social contract for this or some kind of universal basic income. He’s talked about a $14K UBI.
This sounds nice but it doesn’t add up.
I’ll take three examples but I think there are many:
1) Air travel
You might argue this is some kind of luxury but AI is changing fast and aviation changes slowly. In 20 years, we will be flying on the same planes that are in the air today. Theoretically you could take away human flight attendants and pilots (though I don’t think you would want to) but you can’t take away the two biggest costs — fuel and the planes. Those make up more than 50% of your ticket with airport fees another other taxes making up as much as 20%.
AI might save small amounts of fuel with real-time route optimization, predictive maintenance or scheduling but at best, you’re cutting around 20% of costs and more likely just 10%. That could be passed on but flying would be highly inaccessible to 95% of consumers in a UBI world.
2) Materials
Mining and materials are already highly mechanized and energy-intensive industries. There are already many self-driving hauling trucks at mine sites but the wider applications can’t radically change the economics of mining. Yes, there could be some real-time optimization, improved maintenance, labor reduction and process automation but at best you’re reducing costs by 30% because AI doesn’t do much to make it cheaper to break, haul and crush rock.
On the processing side it’s even tougher. Steel and aluminum manufacturing is highly-automated already and incredibly energy-intensive. You can optimize blast furnaces and improve precision but there isn’t a path to ultra-cheap mining. Some might argue that the AI may discover higher-grade deposits but you ultimately still need to drill dozens of core samples to prove a deposit.
3) Manufacturing
This is a big one and it’s wide and varied but most consumer goods manufacturing facilities are either highly automated or located in places where the cost of labor is already extremely low.
In automotive — which has decent wages — factory labor accounts for 6-12% of the retail cost of a car if you add in R&D, engineering, and design labor that might double but it isn’t a path to providing extremely cheap cars. There is some space of optimization in supply chains, quality control and energy but
In these three examples you might save 20-30% of costs (at best) by dramatically cutting back on labor and highly-automating processes with AI but with people losing 80-100% of their salaries, the math doesn’t add up to some kind of age of abundance.
I could go on with examples in agriculture, utilities, healthcare and construction.Three problems AI doesn’t make cheaper are 1) hard physical limits 2) existing automation 3) infrastructure dependencies.
While I strong believe AI will be deflationary, it won’t be enough to sustain the living standards of the millions who will lose their jobs. A 20-30% decline in costs doesn’t fix an 80-100% decline in income. Moreover, any path to a new political consensus will be fraught.
This article was written by Adam Button at www.forexlive.com.