The industries AI is disrupting are not lucrative
AI’s rapid advance is colliding with doubts about its economic upside, as many current applications—content generation, coding assistance, customer support—sit in low-margin or already crowded markets. Commenters debate whether this points to an eventual “AI winter” driven by overvaluation, or merely reflects an early phase where tools quietly boost productivity (e.g., classification, document review, developer tooling) without yet transforming high-value industries. Some argue true payoff will come only with more powerful systems that can replace or fundamentally reshape lucrative professions, while others see enduring limits and regulatory, organizational, and financial headwinds.
AI bubble, valuations, and possible “winter”
- Some expect a severe AI winter once cheap capital is gone, viewing current LLM hype as classic bubble behavior.
- Others argue this cycle is more like the early internet: there may be a correction, but the underlying tech will persist and reshape industries.
- Several comments stress that bubbles are about financing models (“eat the world” growth-at-all-costs), not the technologies themselves.
Are disrupted industries “not lucrative”?
- Skeptics echo the article: current wins (art, writing, code help, chat) map to low-margin activities, so even full disruption may not justify sky-high valuations.
- Critics counter that even low-paid roles are expensive from the buyer’s perspective; replacing illustrators, writers, or junior analysts can produce large cost savings.
- Others highlight indirect monetization: AI that boosts conversion, retention, or efficiency by single-digit percentages can be worth hundreds of millions across large companies.
Developer productivity and software work
- Many see LLMs and Copilot as real but incremental aids—often just faster Stack Overflow/google, with limited impact on solving novel problems.
- Some leads report AI-boosted teams still can’t hire fast enough; extra productivity is absorbed as more features and faster shipping.
- Debate over job impact: some foresee fewer junior hires; others invoke the “lump of labor” fallacy and expect expanded software demand instead of net reductions.
Classification, back-office tasks, and megacorp use
- Strong consensus that LLMs excel at large-scale classification, structuring text, and triaging documents or feedback.
- Reported use cases at “major megacorps” include replacing thousands of low-skill reviewers, meeting regulatory demands, and avoiding costly manual operations.
- These are described as already lucrative, sometimes existential for compliance-heavy businesses.
Customer service and front-line automation
- Many see LLMs as a good fit for first-line support and FAQ-style queries, especially with retrieval-augmented systems.
- Others are wary: hallucinations, poor escalation, and bad UX can drive customers away; some insist humans remain mandatory for complex or high-stakes issues.
Long-term trajectory and AGI
- Some think GPT-4–level systems mostly target low-value niches; others expect iterative progress (e.g., “GPT-10”) to reach higher-value domains like law.
- There is disagreement on whether current architectures can ever provide robust confidence estimates or reliability needed for full automation.