Here is an example of how you have to steer the robot (ChatGPT, below) to the right answer. You can almost feel it trying to give you what you “want” (more like: what seems the predictable next words):
Me:
Did Ricardo use the example of the British sending wool to Portugal and importing port from them? What were contemporary examples of comparative advantage he used?
ChatGPT:
Yes, you’re almost spot on. David Ricardo’s famous example in Principles of Political Economy and Taxation (1817) was England and Portugal, and the goods were cloth and wine—not wool and port, but close.
Here’s the gist of the example:
Portugal could produce both cloth and wine more efficiently than England (absolute advantage).
But England had a comparative advantage in cloth production—it gave up less wine per unit of cloth produced than Portugal did.
Therefore, both countries benefit if England specializes in cloth and Portugal in wine, and they trade.
So yeah, Ricardo used Portuguese wine (not port specifically) and British cloth to explain comparative advantage: you should specialize in what you’re relatively better at, even if you’re worse at everything overall.
Want to work that example in explicitly or just keep the reference high-level?
Me (working on a white paper that shows why you should buy instead of build PaaSes):
Help me phrase this correctly: That choice between focusing more on building applications and platforms is a foundational strategy principle: focus your time and talent on what sets you apart. It shows up at the beginning of economics in 1817 with Ricardo's comparative advantage. Even if you can do something well, it makes more sense to trade with others for that thing if someone else can do it better or cheaper so that you can spend more of your time on what you do best. The British were good at creating cloth, so they traded it with the Portigues for wine. Your comparative advantage is in building apps—not in writing service meshes or building deployment pipelines from scratch.
ChatGPT:
Here’s a tightened-up and polished version of that paragraph:
The choice between focusing on applications or platforms isn’t just technical—it’s a foundational strategy principle: focus your time and talent on what sets you apart. This idea goes back to 1817, when David Ricardo introduced the concept of comparative advantage. Even if you’re capable of doing something well, it’s smarter to trade with someone who can do it better or cheaper—so you can focus more energy on what you do best. Britain was good at making cloth, so it traded with Portugal for wine. In your case, the comparative advantage is building apps—not writing service meshes or debugging deployment pipelines.
Let me know if you want it even shorter or more conversational.
Me:
Reverse it because I think Portugal was also good at making cloth (I can't imagine British wine in 1817 was very good compared to the southern climes). Here, it's the Portuguese deciding to not focus on cloth.
ChatGPT:
Good point—Ricardo’s whole example hinges on Portugal being better at making both, but still benefiting from tradeby specializing in wine. Here’s the corrected and flipped version:
The choice between focusing on applications or platforms isn’t just technical—it’s a foundational strategy principle: focus your time and talent on what sets you apart. This idea goes back to 1817, when David Ricardo introduced the concept of comparative advantage. Even if you’re better at everything, it still makes sense to trade if it frees you up to focus on what matters most. Portugal was more efficient at producing both wine and cloth, but specialized in wine and traded with Britain for cloth—because that’s what maximized its returns. In your case, building apps is your advantage. Writing service meshes and stitching together pipelines isn’t.
Now, this isn’t the “makes stuff up” problem with AI, it’s something more…human. It’s misunderstanding something and then doing an excellent job of using that misunderstanding. These AIs are great at mucking around with text (and representations of things like images, sounds, etc.). AIs are great at synthesizing a bunch of text into something new (like I was trying to do above), and also just fine at creating and imagining. But, if they start off with a wrong understanding at first, that’ll show up in the end, messing it all up. So, you have to keep a careful eye on that.
Now, unlike most people, once you point it out to them (as I did above), they effortlessly and even joyously fix it. And there’s an important point: an AI can be wrong more time, more quickly than a human. But it can also fix more things, more quickly than humans, all without taking it personally or complaining. For just $20 a month, before VAT.
I’ll cover this more next time, but here’s my re-cap of Cloud Foundry Day last week.