TechCrunch has published a glossary aimed at general readers, offering plain-language definitions for the thicket of terminology that has grown up around modern artificial intelligence. The piece targets the growing audience of people who encounter words like "large language model," "hallucination," and "inference" in daily coverage but lack a technical background to parse what they actually mean.
The timing is not accidental. The past two years have produced one of the more dramatic expansions of specialist vocabulary to enter mainstream conversation since the early days of the internet. When a technology moves from research labs to consumer products used by hundreds of millions of people, its internal language tends to follow, often before anyone has bothered to translate it. The same pattern played out with terms like "algorithm," "cloud," and "bandwidth," words that began as precise technical descriptors and eventually became vague shorthand — sometimes losing their meaning entirely in the process.
What makes the current AI vocabulary wave distinctive is the speed at which it is happening and the degree to which the terms carry genuine stakes. "Hallucination," for instance, is not merely colorful jargon. It describes the tendency of large language models to generate confident, fluent, and entirely fabricated information — a behavior with real consequences when the outputs are used in legal filings, medical consultations, or journalism. Understanding what the word actually means, as opposed to treating it as a synonym for "mistake," changes how a person evaluates a tool they might be using at work or at home.
Large language models themselves sit at the center of much of the confusion. The dominant public metaphor — that these systems "think" or "understand" — is almost certainly misleading, yet it is the one that sticks because it is intuitive. The more accurate picture, that an LLM is a statistical system trained to predict likely sequences of tokens based on vast quantities of text, is harder to hold in mind but far more useful for reasoning about what the technology can and cannot do reliably. Glossaries like the one TechCrunch has produced serve a quiet civic function here: they push back, gently, against anthropomorphism that can distort both individual decisions and public policy.
The players with the most to gain from widespread terminological confusion are, this suggests, the companies selling AI products and services. Vague language makes vague promises easier to sustain. When a vendor claims a system is "intelligent" or "reasoning," those words carry different weights depending on whether the audience has any framework for evaluating them. An informed user is a more demanding user, which is why the educational effort represented by accessible glossaries tends to come from journalism rather than from the industry itself.
The consequences of this terminology gap play out across several groups. Policymakers drafting AI regulation are frequently working with imprecise definitions that can make legislation either too narrow to be effective or broad enough to be unworkable. Employers integrating AI tools into workflows may set unrealistic expectations — or dismiss genuinely useful capabilities — because they are operating on metaphor rather than mechanism. And ordinary users, deciding whether to trust an AI-generated summary or a chatbot's answer, are making judgment calls without the conceptual vocabulary to make them well.
For journalists and analysts covering the sector, there is an additional layer of responsibility. Terms of art that circulate uncritically in coverage can harden into conventional wisdom before they have been properly examined. "Artificial general intelligence," to take one example, is treated in some coverage as an imminent and well-defined destination, when in practice there is no consensus definition of what it would mean to have arrived there. The language shapes the story, and the story shapes what the public expects and what regulators feel compelled to address.
What to watch for in the period ahead is whether this kind of accessible, definitional journalism begins to have a measurable effect on public discourse. One signal will be whether political and regulatory debates around AI start to reflect more precise language — whether, for instance, discussions of "bias" in AI systems begin to distinguish between the several quite different phenomena that word can describe. Another signal will be the industry's response: companies have historically moved to co-opt or redefine terminology when precise definitions work against their interests, and there is no particular reason to expect that pattern to change. The glossary is a useful starting point. Whether it becomes a floor or a ceiling for public understanding of a technology that is rapidly becoming infrastructure depends on how much the press, educators, and eventually the companies themselves are willing to sustain the effort.
