How Large Language Models Work, in Plain English
The chatbots answering your questions are prediction engines trained on text. Here is what they actually do, and where they fall short.
The products, platforms, artificial intelligence and policy reshaping how people live and work. This section takes the parts of technology that are usually described in either marketing language or jargon and puts them in plain English — what a large language model is doing when it answers, where “the cloud” physically is, and what end-to-end encryption does and does not protect.
Our test for a technology story is whether a reader comes away able to reason about the thing themselves. That means explaining the trade-offs honestly, including the ones vendors tend to leave out, and being clear about the difference between what a technology can do today, what is demonstrated in research, and what is still a claim.
The chatbots answering your questions are prediction engines trained on text. Here is what they actually do, and where they fall short.
End-to-end encryption keeps the contents of your messages private in transit. It is powerful, and it is also narrower than many people assume.
The cloud is not in the sky. It is rooms full of computers in buildings around the world, and someone else owns them.