Comment by stymaar
7 days ago
> The defining characteristic of a commodity is that it is fungible: a gallon of oil is a gallon of oil; a ton of copper is a ton of copper; a bushel of wheat is a bushel of wheat.
The concept of “commodity” as defined above is a model, a simplified abstract representation of reality, but that does not match the reality perfectly (the map != the territory).
The author claims that a token isn't literally an ideal commodity, but neither is oil or wheat, many factors influence their real value (intrinsic properties, location, available storage at production, expected delivery date, etc.) so that no two gallons of oil in different contracts have the same price.
Is treating “tokens” as a commodity a worse model than treating oil this way? It depends who you ask! I'm pretty sure that a chemist working at a refinery would be more happy to see tokens being felt with like a commodity by his company than if they started viewing crude oil like one.
(Overall, there's way too much economism in that post, and way too few facts, and as a result the argument makes very little sense, the author basically wrote that both OpenAI and Anthropic are drowning in cash right now because compute scarcity means the price must be significantly higher than the marginal cost…)
Yeah I bumped on this as well. Contra the author's claim, the analogy to energy commodities seems very direct to me. Natural gas is not useful in and of itself, what is useful is the energy or aggregates created from it, and those have very different levels of efficiency. Exactly like Sol more efficiently converting tokens into intelligence than Kimi, a combined cycle gas plant converts gas into electricity more efficiently than a simple cycle gas plant. But this does not imply that gas is not a commodity. And both the more efficient and less efficient kinds of plants have large markets; they just target different trade offs.
Edit to add: I think what he's saying is more like "tokens aren't the interesting commodity, 'intelligence' is", which makes more sense. To carry on my gas and electricity analogy, I would say the same thing about gas being the less interesting commodity than electricity, because electricity can be used for a broader set of useful things. But both things are commodities, despite one being an input and the other being an output in this case, and the conversion efficiency is one very important consideration, but not the only one.
"Intelligence" isn't a commodity at all. It doesn't make a lick of sense.
The value of intelligence is that it can solve my specific problems in ways that are satisfying to me. The example the article uses is a CRUD app - but the CRUD app I need isn't fungible with the CRUD app you need! It's not fungible at all, it's a specific solution to a specific problem that may have zero value to anyone else, and certainly cannot be replaced with anyone else's solution to their problems.
If we're comparing electricity to gasoline, then models are cars, and "intelligence" (I disagree that this is what LLMs produce, but whatever) is distance traveled.
Distance traveled is not a commodity. It's the desired outcome.
I think he's creating a definition for an abstract thing here, and using the word "intelligence" to represent that abstract thing.
In your analogy, the "CRUD app" is analogous to the car, but that's not what Thompson is defining to be the unit of "intelligence". He's saying that some number of units of "intelligence" are necessary to create that CRUD app, somewhat analogously to how some number of units of energy are necessary to create a car.
But I agree with you that this concept of a "unit of intelligence" that he's using is probably too abstract to ever be usefully well defined.