Software engineering may no longer be a lifetime career
Source: Hacker News
Why the “skill‑atrophy” argument is weak
I don’t think there’s compelling evidence that using AI makes you less intelligent overall1. However, it seems pretty obvious that using AI to perform a task means you don’t learn as much about performing that task. Some software engineers think this is a decisive argument against the use of AI. Their argument goes something like this:
- Using AI means you don’t learn as much from your work.
- AI‑users thus become less effective engineers over time, as their technical skills atrophy.
- Therefore we shouldn’t use AI in our work.
I don’t necessarily agree with (2). On the one hand, moving from assembly language to C made programmers less effective in some ways and more effective in others. On the other hand, the transition from writing code by hand to using AI is arguably a bigger shift, so who knows? Even if we grant that (2) is correct, this is still a bad argument.
Until around 2024, the best way to learn how to do software engineering was just doing software engineering. That was really lucky for us! It meant that we could parlay a coding hobby into a lucrative career, and that the people who really liked the work would just get better and better over time. However, that was never an immutable fact of what software engineering is—it was just a fortunate coincidence.
It would really suck for software engineers if using AI made us worse at our jobs in the long term (or even at general reasoning, though I still don’t believe that’s true). But we might still be obliged to use it, if it provides enough short‑term benefits, for the same reason that construction workers are obliged to lift heavy objects: because that’s what we’re being paid to do.
Analogy: If you work in construction, you need to lift and carry a series of heavy objects in order to be effective. Lifting heavy objects puts long‑term wear on your back and joints, making you less effective over time. Construction workers don’t say that being a good construction worker means not lifting heavy objects. They say “too bad, that’s the job.”2
If AI does turn out to make you dumber, why can’t we just keep writing code by hand? You can—but you might not be able to earn a salary doing so, for the same reason there aren’t many jobs for carpenters who refuse to use power tools. If the models are good enough, you will simply get out‑competed by engineers willing to trade their long‑term cognitive ability for a short‑term lucrative career.3
I hope that this isn’t true. It would be really unfortunate for software engineers. But it would be even more unfortunate if it were true and we refused to acknowledge it.
A career with a limited lifespan?
The career of a pro athlete has a maximum lifespan of around fifteen years. You have the opportunity to make a lot of money until around your mid‑thirties, at which point your body just can’t keep up with it. A common tragic figure today is the professional athlete who believes the show will go on forever and doesn’t prepare for the day they can’t do it anymore. We may be in the first generation of software engineers in the same position. If so, it’s probably a good idea to plan accordingly.
If you’re thinking “wait, there’s research on this,” you can read my take on the relevant papers:
Mitigations & broader context
Of course, construction workers have layers of techniques for avoiding lifting heavy objects when possible (cranes, dollies, forklifts, etc.). There’s a natural analogy here to a set of techniques for staying mentally engaged that software engineers have yet to discover.
In theory, labor unions could slow this process down (and have forced employers to slow down this race‑to‑the‑bottom in other industries). But I’m pessimistic about tech labor unions for all the usual reasons: the job is too highly paid, you can work (and thus scab) from anywhere on the planet, and so on.
If you liked this post…
- Subscribe to email updates about my new posts:
- Share it on Hacker News:
Related preview
Luddites and burning down AI datacenters
Is it time to start burning down datacenters?
Some people think so. An Indianapolis city council member had his house recently shot up for supporting datacenters, and Sam Altman’s home was fire‑bombed (and then shot shortly afterwards). People from all sides of the argument are sounding the alarm about imminent violence.
The obvious historical comparison is Luddism, the 19th‑century phenomenon where English weavers and knitters destroyed the machines that were automating their work, and (in some cases) killed the machines’ owners. Anti‑AI people are reclaiming the term to describe themselves, and many of the leading lights of the anti‑AI movement (like Brian Merchant or Gavin Mueller) are pushing the narrative that AI threatens human labor.
Footnotes
fbpjLV7epM0fSv_V01QSY5b5TP have written books arguing more or less that the Luddites were right, and we ought to follow their example in order to resist AI automation.
Footnotes
-
See the discussion in the opening paragraph. ↩
-
Construction workers don’t say that being a good construction worker means not lifting heavy objects. They say “too bad, that’s the job.” ↩
-
If the models are good enough, you will simply get out‑competed by engineers willing to trade their long‑term cognitive ability for a short‑term lucrative career. ↩