Artificial IntelligenceArtificial Intelligence

Melanie Mitchell is an AI researcher, and her book, subtitled A Guide for Thinking Humans, is a calm correction to both kinds of hype: the people who say AI is about to match us and the people who say it's a trick. She explains how the main systems actually work, and what they can and can't do.

Her central point is what she calls the barrier of meaning. These systems can be very good at a task without understanding it the way a person does. She shows image classifiers fooled by a few changed pixels no human would notice, and language systems that fail at simple common sense questions a child could answer. She also walks through AI's history of big promises followed by long disappointments, and asks whether this time is really different.

She's fair throughout. She's impressed by what the systems can do, and she's clear about what's still missing.

It's on the shelf because building AetherMem means taking the gap between performance and understanding seriously. A model that looks like it remembers and a model that actually keeps track of what matters are different things. Mitchell is good at making that difference easy to see.