Natural language processing is the field concerned with getting computers to work with human language, understanding, generating or translating text and speech. Unlike a formal programming language, deliberately designed so that every valid statement has exactly one unambiguous meaning, human language wasn't built with that kind of precision in mind at all, and the same sentence can genuinely, legitimately admit multiple different valid readings depending on context, a challenge natural language processing has to confront directly and that more formally precise computing domains simply don't face to nearly the same degree.
Formal languages are built to eliminate exactly the ambiguity natural language embraces
A programming language's syntax is deliberately designed so that any given valid piece of code has exactly one correct parsing and one clearly defined meaning, with genuine ambiguity treated as a design flaw to be eliminated during the language's own development. Human language evolved under an entirely different set of pressures, for flexible, efficient everyday communication between people who can draw on shared context, tone and background knowledge to resolve ambiguity in real time, rather than for the kind of formal precision a computer program requires. This means genuine, unresolved ambiguity is a completely normal, structurally unavoidable feature of human language, not an occasional flaw to be designed away.
Resolving that ambiguity correctly requires context a computer doesn't automatically have
A sentence like "I saw the man with the telescope" genuinely supports two different valid readings, depending on whether the telescope belongs to the observer or the man being observed, and a human reader typically resolves this ambiguity almost instantly using surrounding context and general world knowledge that isn't explicitly stated anywhere in the sentence itself. Getting a computer system to resolve this same kind of ambiguity correctly requires it to draw on comparable contextual and world knowledge, which is precisely why natural language processing has historically been, and in important ways remains, a genuinely harder computational problem than tasks operating over formally precise, unambiguous input.
What we're still unsure about
That human language contains genuine, structurally unavoidable ambiguity, and that resolving it correctly requires drawing on broader context and world knowledge, are well established, foundational observations in linguistics and natural language processing research. What's more genuinely an active, ongoing research question is exactly how completely modern language processing systems have actually solved this ambiguity-resolution problem versus how often they still get it wrong on genuinely difficult or unusual cases, since even the most capable current systems continue to make errors on ambiguous constructions that a human reader would resolve easily using context — researchers continue to actively study and measure exactly where these remaining gaps are, rather than treating the ambiguity problem as fully and completely solved.
This sits inside Natural Language Processing, one of eight topics in AI/ML, one of seven domains in Computer Science, one of seventeen subjects the app can quiz you on.