Kevin Kelly's 2016 AI Predictions, 10 Years Later

This revisits Kevin Kelly’s 2016 forecast about where artificial intelligence would lead by 2026, extracting several AI-focused predictions from his book. Kelly argued that Google’s evolution from a search company to an AI company was deliberate, with every query serving as training data to improve its models, and predicted that transformative AI would sit in the cloud rather than as a humanlike on-device entity. He foresaw an AWS-like utility model delivering industrial-grade intelligence, an oligopoly of a few dominant cloud providers driven by network effects, and a future where conversational access to knowledge is cheap and ubiquitous.

The excerpts also emphasize that most commercial AI would be narrow, specialized software brains rather than general intelligence, and Kelly advocated for freely available models and weights to fuel commerce and science. Specifics include the scale of data and compute powering progress, the transformer breakthrough shortly after his book, the prediction of near-zero marginal cost per answer (nanocents), and the expectation that people will routinely ask hundreds of daily questions to the cloud. The net picture is of pervasive, cloud-hosted conversational and specialist AIs dominated by a handful of providers, with tensions between openness and commercial control.