Here’s a provocative essay: AI is driving down the price of knowledge – universities have to rethink what they offer. The title alone irritated me: it proposes that AI is a competing source of “knowledge” against universities. AI doesn’t generate new knowledge! It can only shuffle, without understanding, the words that have been used to describe knowledge. It’s a serious mistake to conflate what a large language model does with what researchers at a university do — throughout the essay, the professor (an instructor at a business school, no surprise) treats “knowledge” as a fungible product that should be assessed in terms of supply and demand.
For a long time, universities worked off a simple idea: knowledge was scarce. You paid for tuition, showed up to lectures, completed assignments and eventually earned a credential.
That process did two things: it gave you access to knowledge that was hard to find elsewhere, and it signalled to employers you had invested time and effort to master that knowledge.
The model worked because the supply curve for high-quality information sat far to the left, meaning knowledge was scarce and the price – tuition and wage premiums – stayed high.
This is a common error — even our universities market themselves as providers of certificates, rather than knowledge — so I guess I can’t blame the author. He’s just perpetuating a flawed capitalistic perspective on learning. But digging further into the essay, I find abominations. Like this graph, which he claims illustrates “why tuition premiums and graduate wage advantages are now under pressure.”
Hot tip for whenever someone shows you a graph: first, figure out what the axes are.
The Y axis is labeled “Price (tuition/wage premium)”. No units, but OK, I can sort of decipher it. We’re paying a sum of money for college tuition, and after we graduate, we might expect that will translate to a wage increase, so this might represent something like a percent increase in base pay for college graduates over what non-college graduates might get. Fine, I could see doing some kind of statistical analysis of that. But it’s not going to produce a simple number!
For instance, in my cohort of students entering undergraduate education in the 1970s, we all paid roughly the same tuition. Afterwards, though, some of us were English majors, some of us were biologists, and some of us were electrical engineers…and there’s a vast difference in the subsequent earnings of those students. This graph is saying that when knowledge, that is, educated workers, are rare, then an education leads to a premium in wages. I can see that, but I think “price” is going to be far more complicated than is shown.
The X axis though…that’s made up. How do you measure “knowledge accessibility”? What are the units? How is it measured? I’ll have to return to that in a moment.
So there are lines drawn on the graph. One is going down, that’s “demand,” and obviously, going down is bad. The value, or price, of knowledge is declining, a claim that I’m not seeing justified here. Why is it going down? Because the supply is going up, which should be good, since it is going up, but knowledge is some kind of commodity that is being stockpiled, but is being called scarce anyway. Curiously, on this graph, the Price of knowledge is going up as “accessibility” increases, while demand goes down.
I’m not an economist, so the more I puzzle over this graph the more confused I get.
There is also a red dashed line here labeled Supply (AI abundance). Which further confuses me. So supply is scarce if produced by non-AI sources, but abundance if pumped out by an AI?
I was so lost that my next thought was that maybe I should look at the raw data and see how these values were calculated. Hey, look! At the bottom of the graph there was a link to “Get the data,” always a good thing when you are trying to figure out how the interpretations were generated.
Here’s the data. Try not to be overwhelmed.
Seriously, dude? None of that is real data. Those are just the parameters the guy invented to make one line go up and another go down.
I stopped there. That is not an essay worth spending much time on. So maybe AI is not generating knowledge and isn’t the cause of a declining appreciation of the value of knowledge?