# Natural Language Processing Natural Language Processing (NLP) is the subfield of [[Artificial Intelligence]] concerned with getting machines to understand, interpret, and generate human language — text and speech. It covers translation, summarization, sentiment analysis, question answering, speech recognition, and conversational systems. For most of its history NLP was hard and brittle, relying on hand-built grammars and rules ([[Symbolic AI]]). The [[Deep Learning]] era transformed it, and the 2017 [[Transformer]] architecture turned it into the most visible frontier of all AI. The current paradigm is the [[Large Language Model]]: a [[Transformer]] trained on enormous text corpora to predict the next word, which — at sufficient scale — yields systems that can converse, write code, reason through problems, and follow instructions. ChatGPT (built on [[OpenAI]]'s models), [[Anthropic]]'s Claude, and Google's Gemini are LLMs. They are the reason "AI" entered mass public consciousness around 2022. ## Why it matters here NLP is where AI most directly touches the vault's core terrain of **information, persuasion, and speech.** **Disinformation at scale.** LLMs can generate fluent, persuasive, personalized text indefinitely and for almost nothing — a step-change for propaganda, astroturfing, and influence operations. The [[Information Warfare]] and [[Disinformation]] concerns the vault tracks (troll farms, the firehose of falsehood) gain a force multiplier when a single operator can produce millions of unique, on-message posts. Synthetic text also degrades the information commons ("AI slop"). **Content moderation and "bias."** Because LLMs are trained on human text and then fine-tuned ([[RLHF]]) to be "helpful and harmless," their political leanings, refusals, and guardrails became a live culture-war issue — the "woke AI" critique on the right, and concerns about corporate control of speech across the spectrum. What a model will and won't say is now a contested form of editorial power held by a few labs. **Surveillance and analysis.** NLP lets states and platforms analyze speech at population scale — sentiment monitoring, automated censorship, and the parsing of intercepted communications. ## See Also - [[Large Language Model]] — the dominant modern NLP system - [[Transformer]] — the architecture that transformed the field - [[Foundation Model]] — the broader class LLMs belong to - [[RLHF]] — how LLM behavior is shaped after training - [[Disinformation]] · [[Information Warfare]] — the abuse surface