They’re behind the chatbots suddenly everywhere. People treat them as a search engine or a person. They’re neither. An LLM is the world’s most well-read autocomplete.
Large language models are behind the chatbots that suddenly seem to be everywhere, writing emails, answering questions, holding conversations. People react by treating them as either a search engine or a person. They are neither, and the truth is stranger and more useful to understand.
A large language model is a system that predicts the next word, over and over, extremely well. It is the world’s most well-read autocomplete. It is not looking anything up, and it is not thinking. It is guessing what comes next.
Let me prove it. If I say “Sweeeeet Caroline…” a lot of you just answered, automatically, “bom, bom, bom.” Nobody looked that up. You did not reason it out. Your brain has heard the pattern so many times that the next bit just falls out of you.
That is essentially what a large language model does. It has read a staggering amount of text, and from all of it, it has learned which words tend to follow which. When you ask it something, it is not consulting a fact database. It is predicting the most likely next word, then the next, then the next, based on percentages drawn from its training. And because it has read so much, that best guess is usually pretty good, and often exactly right. The model just does it far faster, and across far more topics, than any person could.
An LLM is autocomplete that has read the whole library. It predicts the next word by the odds, not by looking up a fact.
This one insight explains almost everything odd about these tools. They sound confident because a fluent next word is always available. They sometimes state wrong things with a straight face because “most likely next word” and “true” are not the same thing. And they are brilliant at style, tone, and language precisely because that is exactly what predicting the next word rewards. Keep that in mind and you will use them well.
Keep going: What Is an AI Model? explains why there are so many of these, What Is an AI Orchestrator? shows how they team up, and Is It Safe to Tell AI Chatbots Your Secrets? covers what not to type into one.
One short email: the scam making the rounds, one tech word made simple, and one tip to keep you and your family safe. Free, no spam, unsubscribe anytime. Plus a free printable "How to Spot a Scam" sheet for your fridge.
We send one email to confirm it's really you. Your address stays with us and our email service, nobody else. Privacy
The cloud isn't in the sky. It's in giant, humming, refrigerated warehouses full of computers, and they're hungry for electricity and water. Here's what a data center is.
New phones are ditching the little SIM card tray. An eSIM is the same thing, built in and programmed over the air, like a hotel updating your key card without a trip to the front desk.
Click every traffic light. Now do it again. CAPTCHAs are bouncers that keep robots out, and if you keep failing them, the problem is probably your connection, not you.
