AI smells bubbly and there's a reasonable chance it could pop soon, throwing many thousands of AI engineers onto the streets to survive. What are similar alternative jobs and the best way to get ready for them? What's in your Aipocalypse Planner? Even if you disagree with the pessimists, it's always good to be prepared.
I'm more worried about the other scenario - AI continuing to improve steadily and ultimately resulting in significant job displacement among so called "knowledge workers" (including programmers and other engineering types). My concern isn't "what happens when there aren't jobs for AI engineers", it's "what happens when millions of highly-educated, intelligent, and previously highly-paid white collar workers suddenly find that there is no demand for their skills?"
Good SWEs get paid to think and reason about architecture, design, and tradeoffs - not to print out lines of code.
If you are hired for boilerplate coding and no reasoning, you're anyhow screwed career-wise.
If AI/ML reaches the point where it can truly reason in an inductive and dedictive manner comparable to an experienced domain expert, then the Turing Problem has been solved and we're in a new world.
Yes, clearly the impact will hit the entry-level / lower-skilled folks first. But if we project forward not that far into the future, say 7, 10, maybe 15 years, I don't find it implausible to think that we could find ourselves in a world where even fairly senior engineers and architects are impacted. And of course I'm not only thinking about programmers and other IT types: I'm thinking about all forms of "knowledge work" being impacted.
Sci-fi? Maybe. But how many things in our world would have been sci-fi not so very long ago? And I don't see any particular reason to expect AI progress to stop. Although I could see it slowing for a bit at various times. S-curves and all that, innit?.
> I'm thinking about all forms of "knowledge work" being impacted
It's the same story of "all forms" of white collar work.
The big bucks you make in Sales, IB, PE, Medicine, Accounting, Actuary, etc all come from reasoning, and we are still very far away from that level of "general intelligence".
The current market does suck for early career candidates, but that's not yet because of AI taking jobs - it's because of standard outsourcing and roll mergers due to margin pressures.
Fully autonomous reasoning that can truly pass the Turing test, but we aren't there yet.
It feels like a lot of people have fallen for the "Chinese Room trap" that John Searle brought up decades ago in Cognitive Science
Truly autonomous semantic modeling and generation is still very much far away - it basically took us 70 years to barely solve Syntactic modeling, and Pragmatics and Semiotics which are even deeper than Semantics are still at their infancy.
I do think semantic modeling will be cracked within the next decade, but it would still take a generation to solve the quantum leap in pragmatics and semiotics. If a model can reason both pragmatic and semiotic concepts at the same level comparable as indistinguishable as a domain expert, then AGI is de facto solved.
> Fully autonomous reasoning that can truly pass the Turing test, but we aren't there yet.
The thing is, it's not that binary. AI progress will be incremental and gradual and we're already seeing some job impact. So while not guaranteed (this is a complex issue, with lots of variables) it's not at all unreasonable to think that significant job displacement due to AI could happen in the near future (for the sake of argument, I'll say "near future" here means "next 20 years").
There was a similar AI/ML bust in the mid-late 2010s.
If you're truly knowledgeable in your subdomain's technical fundamentals (information retrieval/databases, MLOps/DevOps, GPU Programming/systems programming, etc) you will land a job in other adjacent fields or remain relevant in the AI/ML space.
Algos and Systems Programming are core fundamentals of CS, and weak fundamentals in these two core areas of CS are a major reason SWEs start lagging in their careers.
If you are a Pandas/SKlearn script-kiddie, you're screwed, but for the same reason a front-end or backend dev who doesn't understand architecture or design is screwed as well.
The Perf_Events bug writeup on HN is a great example. A good GPU Programmer/ML Infra Engineer will have that level of Linux Kernel and eBPF knowledge, and could easily pivot into adjacent fields like HFT, Databases, Cloud, etc.
Software is software, the specific nature of it at a point-in-time hardly matters in the overall scope of your career: what happened to the developers building those ANSI POS terminals running in DOS? They learnt new ways to write software, for desktops. Then they learnt new ways to write software, for the web. Etc.
