How to train your data

In 2026, Apple raised prices on iPads and MacBooks by up to $500, and the Apple TV now costs $200. The reason, as David Pierce explains on The Vergecast, is an "era of AI driven shortages of memory and storage." This is the backdrop for a conversation with Alex Reisner of The Atlantic, who has spent years investigating the raw material that makes AI possible: training data. The episode is built around a simple but often overlooked premise: understanding how AI models are trained—and on what—is essential to understanding the technology itself, and the industry is deeply conflicted about letting anyone see how it works.

Why the data matters more than the model

Reisner argues that the training data is arguably the most important aspect of any AI model. "If you train it on 1950s jazz," he says, "that model will be very good at generating music that sounds a lot like 1950s jazz." The model's name—ChatGPT, Claude—is almost irrelevant. "You could make an argument that the right name for a model is actually the description of the data it was trained on, because that is the description of its capabilities."