How accurately did science fiction stories predict AI?

Science fiction stories were largely inaccurate in predicting how AI would actually be developed, particularly regarding the transformer model's success and rapid deployment to billions of users. Ted Chiang's "The Lifecycle of Software Objects" most accurately captured the actual process of teaching, maintaining, and commercializing AI systems, while most pre-2017 works failed to describe key technical breakthroughs like transformer attention mechanisms or the path from chatbots to research-capable AI agents.