Artificial intelligence, judged by what it does rather than what it’s sold as

Most AI coverage falls into two camps. One says the machines are about to think. The other says they’re fancy autocomplete. Neither helps much when an AI summary is the first thing a stranger reads about you, or when a synthetic voice calls your parents sounding like you.

This section looks at artificial intelligence as a representation machine. Large language models predict text from patterns in huge amounts of writing. Image and voice models do the same with pixels and sound. AI agents chain those predictions into actions. In every case the system outputs a model of reality, and the interesting questions are about the gap: what it leaves out, what it invents, and who gets hurt when people treat the output as the world.

We cover deepfakes, synthetic media, AI search and online identity, and we keep a close eye on claims about reasoning and understanding. When a lab announces a breakthrough, we read the paper before the press release. The feed below carries MIT Technology Review, which does the same.