Since last year I’ve had a blog post in my backlog about the weirdness of seeing your profession disappear. I entered Software Engineering from the humanities. After an undergraduate in Comparative Literature, I was cynical about academia and saw Tech as a comparatively meritocratic field where I could actually earn an income. I know a lot of humanities majors who made that transition. Where does this leave me now?
I don’t think the severity of this change has been digested by society yet. I spent a decade building career capital based on an understanding of programming languages, design patterns and system architectures. During that decade new abstractions did arise. In my first job, we had a single managed server hosted in Texas, where I logged in and manually ran database migrations whenever we made a change. Now many companies use serverless cloud architectures. A lot of traditional IT jobs had already fallen by the wayside. 7 years ago, I moved into management expecting leadership to be more resistant to further disruptions in the industry. I saw what had happened to system administrators and expected the increasingly powerful abstractions in cloud systems and web design to do the same to many engineering jobs. I suspected artificial intelligence would accelerate these abstractions, but I didn’t realize it would obviate the micro- and macro-skills of software development altogether. The entire job as it once existed is now automated. It will never again be necessary for an engineer to analyze a method signature or a class design. 90% of my career capital has evaporated. I know competent product managers who can realistically write software that is as good as or better than anything most engineers would build. Often now, the moat in software is an ability to write clear specifications.
Today there is still a need to guide agents, but it feels like cope. We’re witnessing the obliteration of other professions, like mathematics. Terence Tao recently suggested that we will move into the era of “Math 2.0,” where greater emphasis is placed on exposition and navigating research directions. In the AI Safety community, I similarly encounter discussions on the importance of taste. But I don’t really buy that any of this is a sustainable practice. For the next several model generations, humans may very well keep comparative advantages that leave us with an important role in steering and evaluating agents. We don’t yet have examples of AI proposing research questions at the scale of the problems they can solve. But I don’t see any reason to think this last barrier will not fall as well.
I still think superintelligence is the primary risk for us to confront here. But at a personal level it is also hard dealing the day-to-day fallout from AI development! I agree with Nate Silver that we are on track for AI to become the defining issue of the next presidential election. What happens when all the moats dry up? A colleague recently insisted we use AI to draw up a contract. Hiring a lawyer would be a waste of money. The economy has not yet caught up to the fallout. If superintelligence risks are managed well, we may enter a period of creative destruction that leaves us all wealthier and happier than before. But the transition has to be messy.
No major economy I know of has prepared a society-wide retraining program for hordes of white-collar workers whose professions have blurred together into a single “AI Sherpa” job. Certifications still matter. Bureaucracy and necessary accountability will prevent wild, instantaneous obliteration of our careers. AI might do a decent job creating a contract, but I’d still ask a lawyer to review an important deal. But competitive pressures will force change soon. Hungry recent graduates willing to work long hours will outcompete senior employees with personal responsibilities whose domain experience is now irrelevant.
As I navigate a transition into AI Safety, I sometimes feel unappreciated. Don’t these kids care how much time I spent building boring enterprise software? It feels like I have made a mistake by not building my entire career on the exact pillar of knowledge that would remain useful throughout the AI revolution. It feels unfair. That impression of injustice is going to permeate society. A feeling that we did everything right, that we had followed all the instructions and were just about successful, but then the rules changed, and we found out we’d already lost long ago.
I’m fortunate because I’ve already pocketed a successful career. I’m able to focus on my writing and upskilling that might allow me to make some cool contributions in the future. But that doesn’t hold across an entire society. I appreciate we are all trying to take this in stride. But the big changes are already here. My profession is already gone!