Watch on YouTube or below:
Blog: https://blog.tangentfox.com/
Links are kept outside of my YouTube descriptions after YouTube threatened me over linking to the New York Times:
For this video, I have bolded the best sources, with the exception of the bolded warning about the IMF article:
So I was watching this video about how AI is ruining jobs, and this graph immediately stood out to me because I think it needs to be tilted. The point of the graph is to show how AI’s impact on jobs in software development is skewed depending on seniority, with lower seniority positions being most negatively impacted, but it’s presented as there being no impact on senior positions, while there is a very clear inflection point in the middle of 2022 where they chose to normalize this graph to.
Speaking of normalization, it strikes me as weird to see any graph affecting a population not normalized to population growth. When you try to compare things without taking into account the current population (or the population 20 years ago, since this is talking about employment, and the data should probably be compared more to when the current workers were born as a population instead of right now), it can be very misleading. Growth is exaggerated, and stagnation or decreases in a measure can appear to still be growing if their growth is not literally zero, but smaller than the population growth. This is why I tilted the graph, because that is an extremely lazy and inaccurate way to get a slightly better view on what’s actually going on. It’s super inaccurate because how do you know what angle to rotate? And also, rotating it is going to move the data points to different times. It’s a bad way to do things, unless you’re just trying to get the vibes right.
Anyhow, since I’ve already made this little flowchart of things, let’s just take a quick look at the studies’ conclusions ourselves, eh? The Stanford paper is where that graph comes from by the way. Their conclusion is there is substantial decline in early-career jobs, and entry-level work is stagnant entirely. Automation destroys this kind of work. Pay rates are unaffected, which is unsurprising considering entry-level pay hasn’t meaningfully increased since 1970.
(Tangent Zone: These charts are obviously all saying slightly different things, and over different timescales, but they’re telling the same story: Everyone has achieved far more, the economy has grown massively, and yet we are all being paid about the same. This really isn’t taking into account that inflation isn’t a very good measure of how much things actually cost more, and thus even this bleak picture is painting a better perspective than reality.
(Tangent Tangent Zone! This is where my favorite graph comes into play, because it clearly illustrates the problem. As you can see, everything necessary or important for life - food, healthcare, housing, transportation - is more expensive, while fun things are cheaper. The falling costs of unimportant things hides the rising costs of essentials. (On-screen citation: Read the sources I’ve linked to. I’m leaving details out.)))
What was I saying? Oh, right! The concluding statement of this paper is essentially that it’s too soon to see if work is going to be displaced instead of replaced as it has been historically. That’s true, if unsatisfying.
The Harvard paper has a much better conclusion in my opinion because it is more specific and gets to the point that matters: Since 2023, AI has taken away entry-level jobs, which reduces opportunities for advancement. This reduces skill development and wage growth, which increase inequality and reduces upward mobility.
Of course, all of this ignores the fact that measuring jobs is misleading, as it doesn’t take into account how many people have to work multiple jobs. On the bottom, this allows pretending that every job is going to a unique person, inflating the amount of work available, and sweeping under the rug that these jobs are terrible. They don’t pay well, and often require harder work than the jobs they’ve replaced. (This is the gig economy.) It also hides the incestuous relationships between top positions at major companies, where the same person is on the board of several companies, gaining far more money than they have put work into earning.
(Tangent Zone 2: Electric Boogaloo! This graph annoys me to hell and back because this is not how you make a good graph. Combining a percentage change with a literal number is a bad idea, and on top of this, the zero point is in the center on one side, and off the chart entirely on the other. I hate how it completely fails to capture the disconnect between prices and wages by doing those things. Don’t do this. (Start fading out.) If you can’t make two graphs work together in the same place like this, make them two graphs instead of stacking data in a way that doesn’t fit! I mean, Jesus, this is basic design. (End of fade. Should be silent or cut off slightly.))