Loading...
In 1950, MIT mathematician Norbert Wiener published 'The Human Use of Human Beings,' followed in 1964 by 'God and Golem, Inc.' In these works, Wiener described with striking precision the danger of deploying machines optimized for goals humans don't fully control. He wrote explicitly: 'If we use, to achieve our purposes, a mechanical agency with whose operation we cannot interfere effectively... we had better be quite sure that the purpose put into the machine is the purpose which we really desire.' He called the field 'cybernetics' and warned that automation without value alignment could destabilize society in ways no engineer intended. His books were reviewed, discussed briefly, then largely shelved. By the 1990s, philosophy departments had stopped assigning them. Then in 2014, Oxford...
Popular framing: Bostrom invented AI safety thinking with Superintelligence in 2014.
Structural analysis: Availability and recency biases route credit to whoever expressed an idea closest to when it became salient; Bostrom arrived as GPT-2 and AlphaGo made AI feel immediate, while Wiener's identical argument from 1950-1964 felt like historical footnote. The Lindy effect should have pointed the other way — an idea durable across sixty years is more reliable than one freshly minted — but narrative fallacy demands an origin myth, so the field manufactured one. The intellectual amnesia is a system property of how recognition propagates, not a failure of individual scholarship.
The gap matters because it reveals a structural failure in how societies process warnings about emerging technologies: the quality and timing of the warning is far less determinative than the social position of the warner and the cultural readiness of the audience. If alignment researchers treat 2014 as year zero, they will systematically miss insights from earlier technological transitions — nuclear risk, automation anxiety, cybernetics — that contain hard-won wisdom about exactly the dynamics they are trying to model.
