Replies to comments on my "LLMs are eroding my career" post
Large language models are increasingly seen as capable of automating much of routine software development and other knowledge work, raising fears that these roles will be commoditized and that demand for human programmers will shrink. Commenters argue over whether “good enough” AI output will dominate despite quality and reliability concerns, how far and fast AI capabilities are likely to progress, and whether Jevons paradox (more efficiency creating more total demand) will really save tech jobs. Underneath the technical debate runs a deeper anxiety about short‑termist capitalism, eroding job security, and what happens to non‑elite workers if AI substantially replaces human cognitive labor without a corresponding social safety net.
Perceptions of Software, Capitalism, and “Things That Work”
- Several argue customers and investors now prioritize cheap, “good enough” outputs that appear to work over robustness or ethics.
- Others push back, saying poor quality causes constant low-grade user rage, but people feel powerless to hold vendors accountable.
- Public companies, short-termism, and shareholder pressure are blamed for cutting corners and incentivizing grift.
AI, Art, and Resentment in Creative Fields
- Many note intense hostility from artists, tied not just to job loss but to being mocked as replaceable and dismissed by AI boosters.
- Some find AI art/copy “good enough” for commercial uses; others highlight obvious flaws once you know the domain (e.g., surfing).
- A minority admits schadenfreude at previously “arrogant” professions being automated, while others find this attitude cruel and corrosive.
Knowledge, Ability, and What Sets Workers Apart
- Common theme: “mere knowledge” no longer differentiates; ability to ship, solve messy problems, and communicate does.
- Comparisons to plumbers and apprentices: tools (including LLMs) raise the floor, but complex work still needs judgment, responsibility, and trust.
- Some emphasize adaptability and “wayfinding” (integrating conflicting requirements) as future-proof skills; others insist capital ownership ultimately dominates.
Demand for Software and Job Compression
- Debate over whether software demand has an upper bound.
- One side: complexity and automation needs are effectively unbounded; Jevons-like effects increase demand.
- Other side: there are limits to useful complexity and to what people will pay for; many routine dev jobs may commoditize or vanish.
- Some teams report clear “more output because of LLMs” dynamics; others predict high future unemployment for tech workers.
AI Trajectory, Limits, and Hype
- Skeptics note assumptions behind AI maximalism: continued rapid improvement, unlimited capital, and a functioning economy after mass displacement.
- Others counter that even modest continued gains and falling costs could drastically change day-to-day software work within a decade.
- Disagreement over whether current deep-learning approaches can ever “learn good engineering principles” or handle truly novel work.
Societal and Ethical Fears
- Worries about mass unemployment, unrest, and increasing inequality if knowledge work is broadly automated.
- Some foresee elites using AI and agents to concentrate power further, with fewer human checks on “evil software.”
- A few hold out hope for political action or post-scarcity outcomes, but most express anxiety and uncertainty.