# AI Winter A period of reduced funding, interest, and progress in artificial intelligence following a cycle of inflated expectations and subsequent disappointment. There have been two recognized AI winters: the first in the mid-1970s (after early symbolic-AI promises failed to materialize), the second in the late 1980s–early 1990s (after expert systems failed to deliver on commercial promises and the DARPA-funded Strategic Computing Initiative ended). ## The term's current rhetorical function "AI winter" is invoked constantly in mid-2020s AI discourse, but always as a *framing device* rather than a prediction: **Skeptics use it as a warning:** "The last time people said AI would change everything, an AI winter followed." The implication: current enthusiasm is a hype bubble, and the inevitable correction will leave investors, workers, and policymakers stranded. **Industry insiders use it as a contrast:** "Previous AI winters happened because the technology didn't work. This time the technology works." The implication: the transformer revolution, scaling laws, and demonstrated commercial applications (ChatGPT reaching 100M users in two months) mean that the current AI boom has a technical foundation the earlier ones lacked. **Venture capitalists use it as a filter:** "When the AI winter comes, only the real companies will survive." The implication: investment should concentrate in proven winners (OpenAI, Anthropic, the hyperscalers) rather than diffusing across AI startups. The term functions as a collective memory — a reminder that technological enthusiasm is cyclical and that previous AI revolutions ended in disillusionment. Whether the current revolution will follow the same pattern is the defining bet of the 2020s technology economy. ## See Also - [[Scaling Laws]] — the evidence that "this time is different" - [[AGI]] — the unfulfilled promise that triggered previous winters - [[AI Washing]] — the hype behavior that winter-watchers point to