Programming

I admit my read of the article was partly skimming, so maybe they covered this point, but from everything I've seen with LLMs, I'm skeptical their impact is going to change much, unless it's to make things shittier by forcing them where they aren't ready. AI as a whole could change a lot that is hard to predict because AI is kinda synonymous with automation and could be many developments of many different kinds of technologies. But the current crop of AI hype and what it's capable of? Where I see it most taking over is the capitalistic "content churn" industry. For anything that needs to be thinking beyond "cash in and move onto the next one", I don't see how it gets integrated very effectively.

Part of what makes me doubt it is efficiency. Although there are some notable advances in efficiency, such as Deepseek's cost reduction in training, generative AI is overall a resource-heavy technology. Both training and inference are costly (environmentally, in GPUs, etc., not just in price tag). Another point is competence. The more complicated a task is, the easier it is for the AI to make mistakes, some of which only an expert in the related subject matter would pick up on, which makes it a high competence task just to evaluate the AI's results and make sure it isn't doing more harm than good. Another is learning. You could look at the competence example and say, a human in training needs similar evaluation, but the human in training will usually learn from their mistakes, with correction, and not make them as often in the future. The AI won't unless you retrain it and then it is still highly limited due to its statistical and tokenizing nature. Another element is trust. The western market has much more of a vested interest than, say, China, in selling the idea that AI as it is now will work and will integrate and therefore will make a profit; otherwise, its house of cards gold rush investments go to waste and the industry tanks (the fragility of that seen already in how easily Deepseek upset the equilibrium, or lack thereof).

I think programmers and programming as a field is in more danger (or danger of change, depending on how you want to look at it) from capitalists than from generative AI. The field already zipped past a phase where I can still remember reading about someone talking about a fizz buzz example as a test of basic programming competence, to the internet being stuffed to the brim with coding bootcamp stuff and "master algorithms and data structures" doctrine. And that change happened before generative AI. I don't know what the hard numbers are, so I could be deceived on it, but by all appearances, programming became much more saturated via all the "learn to code" stuff, coupled with more companies cutting jobs in general, resulting in it being a field that is significantly harder to get into and harder to stay in. And again, all of that before generative AI.

I don't mean this toward you, Yogthos, of course, but I think there is a certain amount of programmers being in denial about the field being touched by capitalism in general. This sort of unspoken belief that because programming is so important, the trend will just sort of continue that way and it will continue to be a lucrative and cozy ivory tower to hang out in. But that won't stop capitalists from trying to reduce payroll as much as possible, whether it truly makes rational sense or not.