AI and the Web Developer Job Market
We have the worst job environment for tech in over two decades and that’s with the “AI” bubble in full force. If that bubble pops hard before the job market recovers, the repercussions to the tech industry will likely eclipse the dot-com crash.
[…]
Not every CEO is predicting the end of programming as a profession. GitHub’s CEO instead thinks that programmers will be the primary beneficiaries of the introduction and improvement of LLM tools for coding.
[…]
This world-view assumes that the purpose of software development is the productive creation of successful, defect-free, software projects. LLMs would increase productivity.
The alternate world-view, one that I think is much more common among modern management, is that the purpose of software development is churn.
Via Kyle Hughes:
The overwhelming thing I agree with is: the observable disconnect between software quality and business outcomes is what is ripe for exploitation in the short-to-mid-term with LLMs. We don’t need to agree on the efficacy of LLMs to agree on this. If management could reduce quality and cost by an order of magnitude while maintaining or slightly sinking outcomes, wouldn’t they?
I think Apple platform developers are more likely to be caught off-guard by the viability of business success on top of bad software. I think we assume the success of iOS in particular is due to software quality but I have always suspected it was a secondary factor, if relevant at all. We could all collectively uphold that illusion because we were the self-selected people that cared and did the work. It’s hard to tell if it’s just our pet concern.
Previously:
4 Comments RSS · Twitter · Mastodon
Maybe it’s because I’ve spent the last 4 months working on a labyrinthine web project that requires human sensibilities at every stop, but I just don’t see the bloodbath coming (or lasting if it does). The sum total of all knowledge has been fed into a learning model that can make decent python scripts. Congrats. But the same last-mile problem that plagues self-driving cars will do the same here. LLMs can code, but only so elaborately and intelligently, and that last mile takes a long, long time to traverse.
I see the bloodbath coming. In fact, I think its first wave is already here.
I hear of efforts to replace all a company's custom solutions with "standard" off the shelf tools that don't even do what is needed. But, thanks to these new "AI" based tools, PowerBI for instance, management feel empowered. They can cobble together "dashboards" to their hearts' content. If management or any intern can do this, why pay so much for people who know how to code?
Of course, once you get rid of people who can code, you no longer understand why databases need to be designed not to mix different concerns and your data becomes less and less useful... For a while no one notices, until eventually you have a pile of poorly thought out unstructured data. Cue, the next AI snake-oil salesman, who assures the CEO that this can be dealt with using the latest "AI".
Like you, @Billok, I think this is destined to fail, but that it may take years to shake out. The longer it takes, the more likely the end result will be that many software jobs will be outsourced instead of brought back because the pool of competence left by that point has been so depleted, and there's no longer enough profit margin left to pay the salaries needed to convince the intelligent to invest in software development as a career.
Recall what Tim Cook said about the reason he doesn't think Apple could still manufacture their wares in the US: there aren't enough skilled people who know how to make stuff in the US. https://www.cbsnews.com/news/60-minutes-apple-tim-cook-charlie-rose/
The same goes for software, once everyone has been replaced by the cheap and cheerful clueless wielding LLMs, those who found another career path will probably stick with it. Even there, LLMs may also wreck their havoc, further reducing opportunities for the intelligent. Since 2/3 of the good techies in the US came from somewhere else, they may return to their countries of origin if things become sufficiently bad. Since corporations have no loyalty to their home countries, they may simply move somewhere else if that's where the "talent" they need is.
So I think the risk is real, not because it makes sense, but because it benefits the people at the top of the organization, even if it doesn't benefit the organization. https://rationalwiki.org/wiki/Iron_law_of_institutions And, as Boeing has demonstrated, management can ignore engineering for a very long time before people start noticing the consequences. https://youtube.com/watch?v=rvkEpstd9os
@Billyok I think the article does not disagree with you. The point is that management believes that LLMs can replace a large chunk of the work force and they can get away with it because business success is often not correlated to software quality. Especially in web tech, given that a lot of web sites are already dysfunctional and people have become insensitive to it.
As someone who has been doing "web tech" for over two decades I wonder what people think f when they hear web tech (and say that it's dysfunctional)
The verge?
Gmail and YouTube?
Figma?
I agree that almost all adsuported websites suck balls without a thick layer of adblockers, but I disagree that webapps are crap because web. Agai, I'm biased.
I do agree with Baldur that it doesn't matter, since the people singing AIs praise
1: Are important to listen to since they are the ones calling the shots
2: Don't give a fuck about quality because of capitalism
I was there wen the -com bubble burst. What happened was me and my friends got fired from cozy jobs and started our own companies. Then when the economic crisis hit and we could get paid for building entire websites/webapps in flash, we had to rethink and build smarter.
Maybe we will see similar things here. Maybe the capitalist wet dream of "If the big guys become stupid, the little ones will flourish" will happen. The DMA will help in the EU.
