Ask HN: What do you expect from the next 5 years of AI?

A community discussion asking what to expect from the next five years of AI collects sharp, often pessimistic takes: many contributors warn that a flood of LLM-generated content will saturate online and workplace communication, increasing noise and making routine "bullshit" easier and more prolific. Several comments highlight the growing ability of models to fabricate convincing falsehoods, while others foresee social fragmentation, even violence, alongside technical progress such as new theorems. Some participants call out the forum’s own biases, noting cynicism about technology among respondents.

Predictions cluster around two structural shifts. One camp expects Western regulation and commercial gating - approved public models for general users and powerful systems reserved for corporations and verified individuals, with open-source competition diminished. Another foresees decentralization of capability: small, local personal models achieving near-frontier reliability while major labs pivot to serving governments and large companies, creating an AI bifurcation. Commenters cite recent examples of restricted access to advanced models (Claude/Opus 4.6) as a preview, and a few expect economic downsides such as lost value or wasted investment.