What happens in a world of AI engineers?
Of artisans and guilds, and the siphon and the forge.
Will we get to AGI and ASI, will our creations surpass us, will they become AI super-scientists discovering new secrets of the universe? I dunno. (Or, taking the long view, I dunno when.) But: will we get AI super-engineers? That answer is, “yeah, probably.” At least for software.
…So what happens then?
The Software Super-Engineers
Traditionally at this point one argues either “AI will destroy all jobs, driving us into a corporate dystopian precariat hellscape!” or “AI will create endless new jobs, ushering us into a fully automated luxury utopia!” In reality, every change brings both costs and benefits. To my mind the likely outcome is twofold:
Fewer people will produce much more, and better, professional software.
(Basic-ish supply and demand. AI will vastly increase the supply of software, which will induce more demand … while the number of devs will diminish.)Many more people will incorporate custom AI-written software into what they do.
Is it bad that there will be fewer software jobs? I don’t know; is it bad that there are many fewer farming jobs than a century ago? Nope. Fewer manufacturing jobs than half a century ago? …OK, that’s more complicated, but I’ve worked in a factory, and I promise you, service jobs are better. Are there shitty service jobs? There sure are! But that doesn’t mean manufacturing jobs are inherently good, whether you’re talking about making furniture or software.
In an ideal world, remember, we would all choose our own jobs, most of which would probably be economically unviable passion projects and/or art, funded by basic income from the dividends of AI software, robot agriculture, etc. Those dreadfully concerned that poor people need jobs for meaning never seem quite so concerned about the idle rich who swan from Burning Man art projects to a month off in Medellín to a new rock band. Obviously the solution is for everyone to be idle rich … while some continue to work like demons to reverse climate change, build Marsbase, and make the Sahel bloom. AI engineers could be a strong step in that direction.
…That said, when I consider the transition, I feel uneasy.
The Guild
Software jobs had a special place in the world. (And whooo it feels weird to be writing about them in the past tense, even speculatively, even as a card-carrying science fiction author.) They were, albeit accidentally, a check and balance on a surprising number of iniquitous forces.
Software became a profession more lucrative than the law (yes, really); much more lucrative than architecture, accounting, or most other forms of engineering; and far, far more independent than academia — all without any formal gatekeeping. Into the 2020s it remained possible, admittedly not likely but certainly possible, to parlay six months in a software bootcamp into an illustrious and highly paid software career, or to become a millionaire bug hunter aged 25. Try that in medicine!
(I’d venture to guess that most software people know comparable success stories personally, btw. One of mine is an ex-colleague who didn’t graduate high school but went on to have a very successful software career at, among others, Meta and Google.)
Software was the last field where you could be, simultaneously:
an autodidact
highly successful without any diploma or licensing
wealthy and in-demand enough to have considerable negotiating power
an artisan contractor, charting your own course from one job/client to another
not unusual — in that all of the above were fairly common, and there were many millions of software engineers worldwide.
It was also a profession available to almost anyone in the world, and, as such, good for almost every nation. At my old consulting company, where we had a mix of “onshore,” “European,” and “offshore” engineers, over the years we funneled millions of dollars from US clients to super-smart people in India, Brazil, etc., all cheap by American standards, all making far more than their nation’s average income.
It was also, quietly, a hugely influential profession. From its ranks came Silicon Valley’s tycoons, the captains of America’s last effective engine of change. And, in politically paralyzed America, those same ranks acted as a check on those same tycoons; for better or worse, one of the very few constituencies that Big Tech had to warily respect was its own employees.
Ultimately, you could argue that software was so different from other professions that it wasn’t a profession at all, but really more of a loose and anarchic guild.
Is a world where that guild is much diminished a better place?
Its remaining members might not be diminished, to be clear. In a future of AI super-engineers, their human conductors might be the new surgeons. But a smaller, more elite software guild will no longer be a globally available pathway out of poverty. It will no longer have a base broad enough to simultaneously include both visionary leaders and rank-and-file reactionary opposition … a tension that was, to my mind, mostly healthy.
Most of all, the idea of such a world raises the question of what new jobs will be created in this new era — and whether they’ll be anywhere near as open, lucrative, independent, and common as software engineering used to be.
The Siphon And The Forge
Long ago, in my guise as a TechCrunch columnist, I wrote a piece called “The Siphon And The Forge,” about how the economy had bifurcated into “forge” industries, which tried to generate new wealth, and “siphons” which simply tried to reappropriate previously generated wealth. This is of course not a particularly original observation:
(Sobotka was half wrong, mind you. The US manufactures more than ever … but those forges employ many fewer people, especially as a fraction of the population.)
Forges are better than siphons. But as nations become wealthier, siphoning increasingly becomes the coldly rational choice — see also the infamous “resource curse” — and, also, industries which used to be forges morph into siphons en masse. The financial industry and stock market are key to capital, credit, price discovery, a healthy economy, etc.; but profiting by settling trades in microseconds instead of seconds, or selling CDOs? That’s a siphon. Advertising was a good and important forge, people should know what’s available; but the online adware biz? Total siphon.
I don’t think most people bother to make the distinction between siphon and forge at all, even locally, much less think about the ramifications. For my startup I’ve been talking to lots of people about what they need from their engineering organization. We expected much of our market to be businesspeople who want to know what’s really happening in engineering.
…It turns out most people on the business side have very little interest in engineering. They want to know when engineering will deliver, and if they’re telling the truth — our product was recently memorably described as “a bullshit detector,” which, I mean, I guess it can be, but it’s intended for less pathological organizations/situations — because engineering is often viewed purely as a cost center. Or, if you will, as a siphon. The idea that a company is investing in engineering, to forge value … has sometimes never occurred to MBAs.
(When we talk to engineering leaders who manage multiple projects, by contrast, they get it quickly, and react “oh yes, I want this, to get independent objective analysis of the forge, and thereby align everyone on what’s being built and how.” …OK, fine, they don’t actually use the word ‘forge,’ but they might as well.)
You see where I’m going with this. Software engineering is not just a but arguably the forge job of this era. When AI supplants it … will the subsequent newly created jobs be forges, or siphons? Or in fields that become siphons, as finance and advertising did? Art and entertainment and documentary-making are forges, and important ones. …but when everyone wants to be a YouTube vlogger, and they’re all competing for the same audience in the same ways, uhhhh, maybe not so much?
A brighter future is visible too. Maybe the next guilds will be “climate repairperson” and “Marsbase engineer” and “ephemeral installation artist,” all mutually inspired, all empowered and financed by AI productivity. Even if everyone becomes a life coach for everyone else’s anxieties in a fully automated AI-enabled luxury utopia … I guess that’s no bad thing, he said reluctantly.
But I worry about the trend of forges turning to siphons, as people and companies and nations grow wealthier — while the forge jobs of this era, artist and engineer, seem first in line for AI transformation. Again, there are many ways the coming transition could be wonderful! …I’d just like to feel like we’re actively trying to steer its course more towards the forge.