The ELIZA code has been found. Did Weizenbaum invent her secrets?
mit researchers recovered the original source code for the 1960s eliza chatbot, revealing new technical details behind the program's famous interactions. the analysis of this missing code clarifies the mechanics behind the eliza effect and how users attribute intelligence to simple text scripts.
after 60 years, the original source code for the eliza chatbot has been recovered from the mit archives, contradicting long-held beliefs about its capabilities. the new analysis in 'inventing eliza' reveals that weizenbaum likely staged or heavily edited the famous dialogs, rather than capturing spontaneous conversations. these findings expose the mechanics behind the eliza effect, where users mistakenly attribute intelligence and empathy to simple keyword-matching scripts. the discovery forces a re-evaluation of the anthropomorphic claims surrounding modern chatbots like chatgpt.