AI Hype Is Cooling – New survey

Original Article ↗ Hacker News Discussion ↗

Startup adoption experiments & tooling costs

  • Some small companies are buying “every AI add‑on” (ChatGPT Enterprise, Gemini, Copilot, Slack/Notion AI, etc.) for a year to see what sticks.
  • Reported spend: $100–$200 per employee per month, i.e., tens of thousands per year — comparable to a salary.
  • Expectation: 2024 is experimentation; 2025 will cut underused tools. Many doubt smaller firms can afford this level of trial-and-error.

Enterprise usage, security, and access

  • Some organizations block external AI tools over data leakage, confidentiality, and compliance fears.
  • Counterpoints: companies already trust many third parties with data; AI vendors offer training opt‑outs and business data agreements.
  • Suggested middle ground: clear security criteria, local/hosted models, or tightly scoped integrations.

Hype cycle, progress, and “AI winter” worries

  • Many frame the current mood as classic hype-cycle cooling: expectations reset after hands‑on experience.
  • Some say frontier models may be hitting diminishing returns; rumors of stalled breakthroughs and underwhelming new models are cited.
  • Others note deep learning has scaled for a decade and expect more gains, tied to compute and architecture advances.

Where AI is delivering value

  • Strong use cases mentioned: coding assistants, internal search over documents, language translation, medical transcription, RAG-based support tools, and medical imaging enhancement.
  • Some find general chatbots most useful for idea exploration; many derivative “AI features” feel gimmicky or harmful to productivity.

Incumbents, platforms, and API wrappers

  • Widely shared view: big platform players (Microsoft, Google, Adobe, Nvidia) are best positioned to win via deep integration and distribution.
  • Skepticism that “API wrapper” startups and meeting-bot tools will survive once incumbents bundle similar features.
  • Some note AI is still largely “integration work”: real value comes from grounding models in company data and workflows.

Economic, social, and ethical concerns

  • Debate over disruption examples like Chegg: some see useful creative destruction; others see VC-subsidized dumping destroying viable, profitable businesses and jobs.
  • Concerns about massive resource use (energy, water, silicon) and “wasteful” AI features that don’t solve real problems.
  • Workers often hide AI use due to fears of seeming lazy, incompetent, replaceable, or sloppy given hallucinations and plagiarism risks.