{"id":160,"date":"2026-11-18T14:36:18","date_gmt":"2026-11-18T14:36:18","guid":{"rendered":"https:\/\/www.kerloys.com\/?p=160"},"modified":"2026-11-18T18:37:54","modified_gmt":"2026-11-18T14:37:54","slug":"what-is-an-llm","status":"publish","type":"post","link":"https:\/\/www.kerloys.com\/index.php\/2026\/11\/18\/what-is-an-llm\/","title":{"rendered":"What Is an LLM? A Plain-English Guide to the Brains Behind ChatGPT"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">If you&rsquo;ve used ChatGPT, Gemini or Claude, you&rsquo;ve already talked to an LLM. You type a question, and a few seconds later a fluent, confident answer appears &mdash; sometimes brilliant, occasionally wrong. So what&rsquo;s actually going on behind that little text box?<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This is the first article in <strong>AI Explained<\/strong>, a five-part series where I break down the AI buzzwords without the jargon. No maths, no code &mdash; just how it works and what it means for you.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">The short answer<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">An <strong>LLM &mdash; a Large Language Model &mdash;<\/strong> is a computer program that has read an enormous amount of text and learned to predict which words are most likely to come next.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">That&rsquo;s it. Everything else you&rsquo;ve seen it do &mdash; answering questions, writing emails, summarising reports, even writing code &mdash; grows out of that one skill, done extraordinarily well.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Breaking down the name<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Large<\/strong> &mdash; it learns from a vast amount of text: books, articles, websites, code. Far more than any person could read in many lifetimes. The model itself is huge too, with billions of internal &ldquo;dials&rdquo; (called <em>parameters<\/em>) that store what it learned.<\/li>\n\n\n<li><strong>Language<\/strong> &mdash; its whole world is text. Words go in, words come out. Newer models can also look at images or listen to audio, but language is the core.<\/li>\n\n\n<li><strong>Model<\/strong> &mdash; a simplified mathematical picture of something. A weather model predicts tomorrow&rsquo;s weather; a language model predicts the next words in a sentence.<\/li>\n\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">How does it learn? Think &ldquo;autocomplete on steroids&rdquo;<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Your phone keyboard suggests the next word as you type. Type &ldquo;See you&rdquo; and it offers &ldquo;tomorrow&rdquo; or &ldquo;soon&rdquo;. It learned that from the way people usually write.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">An LLM does the same thing at a vastly bigger scale. During training, it&rsquo;s shown a piece of text with the next word hidden, makes a guess, checks the real answer and adjusts itself very slightly. Repeat that billions and billions of times across a huge library of text, and something remarkable happens: to get good at predicting the next word, the model has to pick up grammar, facts, writing styles, reasoning patterns &mdash; even a bit of humour.<\/p>\n\n\n\n<blockquote class=\"wp-block-quote is-layout-flow wp-block-quote-is-layout-flow\">\n<p class=\"wp-block-paragraph\">To finish the sentence &ldquo;The capital of France is&hellip;&rdquo;, you need to know some geography. To finish &ldquo;If the website is down, the first thing to check is&hellip;&rdquo;, you need to know some IT. Predicting words well forces the model to absorb a lot of knowledge along the way.<\/p>\n<\/blockquote>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>A quick word on &ldquo;tokens&rdquo;.<\/strong> Strictly speaking, LLMs don&rsquo;t read whole words but <em>tokens<\/em> &mdash; small chunks of text, often a word or part of one. &ldquo;Cinephile&rdquo; might be split into &ldquo;cine&rdquo; and &ldquo;phile&rdquo;. You&rsquo;ll see the word <em>token<\/em> whenever people talk about AI pricing and limits; just think of it as roughly three-quarters of a word.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">From word-predictor to helpful assistant<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">A model that has only learned to predict text is like a brilliant but unruly student: it knows a lot, but doesn&rsquo;t know how to behave in a conversation. So there&rsquo;s a second stage of training. People write example questions with good answers, and rate the model&rsquo;s replies as helpful or unhelpful, safe or unsafe. The model is then tuned towards the kind of answers people prefer.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">That&rsquo;s what turns a raw text predictor into the polite, instruction-following assistant you chat with.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">So&hellip; does it actually understand?<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">This is where even experts disagree, so it&rsquo;s worth being honest about it. An LLM doesn&rsquo;t think the way you do. It has no experiences, no feelings and no memory of yesterday&rsquo;s chat unless it&rsquo;s given one. But it has clearly learned deep patterns about how ideas fit together &mdash; which is why it can explain a concept three different ways, or spot the bug in your code.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">A useful way to think about it: <strong>an LLM is like a very well-read colleague who answers instantly, never gets tired, and is sometimes confidently wrong.<\/strong><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">What LLMs are great at<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Drafting and rewriting<\/strong> &mdash; emails, reports, social posts, cover letters.<\/li>\n\n\n<li><strong>Summarising<\/strong> &mdash; turning a 30-page document into five bullet points.<\/li>\n\n\n<li><strong>Explaining<\/strong> &mdash; &ldquo;explain this like I&rsquo;m twelve&rdquo; works surprisingly well.<\/li>\n\n\n<li><strong>Brainstorming<\/strong> &mdash; titles, gift ideas, project plans, names for your Wi-Fi.<\/li>\n\n\n<li><strong>Translating and changing tone<\/strong> &mdash; formal to friendly, English to Malayalam and back.<\/li>\n\n\n<li><strong>Coding help<\/strong> &mdash; writing, explaining and fixing code.<\/li>\n\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">Where they fall short<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>They make things up.<\/strong> When the model doesn&rsquo;t know, it may still produce a fluent, confident answer &mdash; including invented facts, quotes or references. This is called a <em>hallucination<\/em>. Remember: it&rsquo;s predicting what a good answer <em>looks like<\/em>, not looking it up in a database.<\/li>\n\n\n<li><strong>They have a knowledge cut-off.<\/strong> A model only knows what was in its training data, which stops at a certain date. Unless it&rsquo;s connected to web search or your own documents, it won&rsquo;t know about last week&rsquo;s news.<\/li>\n\n\n<li><strong>Numbers and exact details can slip.<\/strong> Arithmetic, dates and precise figures are getting better, but double-check anything that matters.<\/li>\n\n\n<li><strong>No real-world judgement.<\/strong> It doesn&rsquo;t know your company, your situation or what&rsquo;s at stake &mdash; unless you tell it.<\/li>\n\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">Many of these weaknesses are exactly what the rest of this series is about. <strong>RAG<\/strong> gives the model an &ldquo;open book&rdquo; of trusted documents so it makes fewer things up. <strong>Agentic AI<\/strong> lets it use tools &mdash; search, calculators, your calendar &mdash; instead of relying on memory alone.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">A security person&rsquo;s advice before you dive in<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">I work in cybersecurity, so I can&rsquo;t write about AI without a few words of caution:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Don&rsquo;t paste secrets.<\/strong> Passwords, customer data, confidential documents and internal code don&rsquo;t belong in a public AI tool unless your organisation has approved it. Check whether the tool uses your conversations for training &mdash; and how to switch that off.<\/li>\n\n\n<li><strong>Verify before you trust.<\/strong> Treat AI answers like advice from a knowledgeable stranger: useful, but check anything important, especially legal, medical, financial or security advice.<\/li>\n\n\n<li><strong>Watch out for AI-powered scams.<\/strong> The same technology writes very convincing phishing emails and fake messages. Bad grammar is no longer a reliable warning sign.<\/li>\n\n\n<li><strong>You&rsquo;re still responsible.<\/strong> If you send it, publish it or run it, it&rsquo;s yours &mdash; AI-written or not.<\/li>\n\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">The names you&rsquo;ll hear<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Some of the best-known LLM families and the companies behind them:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>GPT<\/strong> &mdash; OpenAI (the models behind ChatGPT)<\/li>\n\n\n<li><strong>Claude<\/strong> &mdash; Anthropic<\/li>\n\n\n<li><strong>Gemini<\/strong> &mdash; Google<\/li>\n\n\n<li><strong>Llama<\/strong> &mdash; Meta (openly available for others to build on)<\/li>\n\n\n<li><strong>Mistral<\/strong> &mdash; Mistral AI, from France<\/li>\n\n<\/ul>\n\n\n\n<p class=\"wp-block-paragraph\">They differ in strengths, price and privacy terms, but they all run on the same basic idea described above.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Key takeaways<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li>An LLM is a program trained on huge amounts of text to <strong>predict the next word<\/strong> &mdash; and that one skill turns out to be incredibly powerful.<\/li>\n\n\n<li>Extra training with human feedback turns it into a <strong>helpful assistant<\/strong>.<\/li>\n\n\n<li>It&rsquo;s brilliant at drafting, summarising and explaining &mdash; but it can be <strong>confidently wrong<\/strong>.<\/li>\n\n\n<li><strong>Keep sensitive data out<\/strong>, and verify anything that matters.<\/li>\n\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\">Next in the series<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\"><strong><a href=\"https:\/\/www.kerloys.com\/index.php\/2026\/11\/18\/prompting-101\/\">Part 2: Prompting 101<\/a><\/strong> &mdash; how to talk to AI to get far better answers. After that: <strong>What is RAG?<\/strong>, <strong>What is Agentic AI?<\/strong>, and <strong>Using AI safely<\/strong> at work and at home.<\/p>\n\n\n","protected":false},"excerpt":{"rendered":"<p>Large Language Models power ChatGPT, Claude and Gemini. Here&rsquo;s what they are, how they learn, what they&rsquo;re good (and bad) at &mdash; in plain English, no maths required. Part 1 of AI Explained.<\/p>\n","protected":false},"author":1,"featured_media":186,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_kad_post_transparent":"","_kad_post_title":"","_kad_post_layout":"","_kad_post_sidebar_id":"","_kad_post_content_style":"","_kad_post_vertical_padding":"","_kad_post_feature":"","_kad_post_feature_position":"","_kad_post_header":false,"_kad_post_footer":false,"_kad_post_classname":"","footnotes":""},"categories":[36],"tags":[40,42,43,41],"class_list":["post-160","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai","tag-ai-explained","tag-chatgpt","tag-generative-ai","tag-llm"],"_links":{"self":[{"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/posts\/160","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/comments?post=160"}],"version-history":[{"count":2,"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/posts\/160\/revisions"}],"predecessor-version":[{"id":164,"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/posts\/160\/revisions\/164"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/media\/186"}],"wp:attachment":[{"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/media?parent=160"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/categories?post=160"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kerloys.com\/index.php\/wp-json\/wp\/v2\/tags?post=160"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}