Inventing ELIZA Recovers 60-Year-Old Source Code, Revealing New Chatbot Dialogs
Updated
Updated · WIRED · Jul 14
Inventing ELIZA Recovers 60-Year-Old Source Code, Revealing New Chatbot Dialogs
1 articles · Updated · WIRED · Jul 14
Summary
"Inventing ELIZA" says it has recovered the original ELIZA source code from MIT Archives, filling a long-missing gap in the history of the 1960s chatbot.
The book argues ELIZA was not just the famous DOCTOR script: the recovered material shows multiple program versions, technical innovations and newly uncovered dialogs for other personas.
Those findings recast Joseph Weizenbaum’s system as a broader experiment in how users project understanding onto machines — the dynamic later labeled the "ELIZA effect."
The authors link that history to today’s AI, arguing modern chatbot interfaces still obscure underlying machinery and labor much as ELIZA exposed the risks of mistaking scripted responses for intelligence.
With 'AI slop' flooding the internet, how can we find truth in the 'textpocalypse' we were warned about?
Is your AI companion a helpful friend, or is it 'chatbait' designed to exploit your emotions for profit?
Restoring ELIZA: How the 2025 Recovery of the Original 1966 Code and "Teaching Mode" Reshapes AI Understanding
Overview
The ELIZA Reanimated Project began after a 1966 magnetic tape labeled 'ELIZA' was discovered at MIT CSAIL in late 2023. A team of computer historians and software archaeologists launched the project in early 2024, facing major technical challenges due to the tape’s degradation and the obsolete MAD-SLIP programming language. Using specialized recovery techniques and building a custom IBM 7094 emulator, they successfully restored the original ELIZA code by mid-2025. This effort not only preserved a key piece of AI history but also revealed new features and deeper insights into Joseph Weizenbaum’s pioneering work in human-computer interaction.