LLMs Aren't Inevitable

The piece argues that large language models are not an unavoidable technological destiny by defining what "inevitable" actually means in practical, social terms. Inevitability requires a technology to perform tasks at least as well as alternatives, serve socially valuable and scarce functions, deliver an overwhelming advantage in volume or quality, and be difficult for non-adopters to neutralize. Historical comparisons - gunpowder and cannon as genuinely inevitable in medieval Europe, versus the magnetic compass as powerful but avoidable - illustrate how social structures and use-cases determine whether a technology forces universal adoption.

Applying those criteria to modern LLMs, the assessment is negative. LLMs can replicate large swathes of mediocre knowledge work and therefore substitute for low-value tasks, but they cannot reach top human standards or reliably handle core bottlenecks that matter most to organizations. Productivity gains are modest rather than revolutionary, there is no clear market actor crushing competitors via LLMs, and social backlash against low-quality AI output creates practical neutralization pathways. Because LLMs fail the overwhelming-advantage and un-neutralizability tests, widespread adoption is contingent and avoidable rather than inevitable.