Moving through the history of Artificial Intelligence, I approach the era when the ideas began to take shape, ready for direct interaction with people. Today, I want to talk about the first one to speak to humans, the one that captured widespread attention, and gave rise to an entire psychological phenomenon.
I will tell you about ELIZA.
>the_originsCreated between 1964 and 1967 by Joseph Weizenbaum, ELIZA will always be remembered as the world’s first chatbot. An interesting observation: Weizenbaum didn’t set out to actually invent a chatbot. He aimed to study human-machine conversation and important cognitive processes like interpretation and misinterpretation. However, despite his intentions, ELIZA gained popularity due to its timely creation and widespread adoption.
>the_processWritten in MAD-SLIP, a language developed by Weizenbaum himself, ELIZA simulated conversations via pattern matching. Its most famous script, DOCTOR, mimicked a Rogerian psychotherapist: after Identifying keywords in user input, it responded with pre-written templates, often rephrasing statements as questions.
As simple as it is by modern measures, many believed ELIZA could actually understand them. After finding this out, Weizenbaum was surprised and alarmed by how little it took for users to anthropomorphize his creation.
>the_legacyELIZA’s success stemmed from its role in early AI research, intersecting with key developments in computing history. Still, while persisting in today’s ChatGPT and Siri, the story of ELIZA also highlights critical lessons:
- AI has limits, and illusions of understanding can be misleading.
- Critical thinking is essential in humans, especially as AI grows more advanced.
Disturbed by people’s emotional attachment to ELIZA, Joseph Weizenbaum later wrote "Computer Power and Human Reason", arguing against anthropomorphizing machines, later known as the ELIZA effect — a phenomenon of attributing human traits to simple programs.
>the_resurrectionIn 2024, researchers brought ELIZA to life from long-lost code found in MIT archives. The original 420-line program, discovered in 2021, was restored by cleaning and debugging the code and emulating 1960s hardware. Although, in order not to "ruin the authenticity of the artifact", the team decided to leave a single bug unfixed, the one making a program crash if a person entered a number; they compared it to "fixing a mis-stroke in the Mona Lisa."
#AIindustryevents #upcomingartificialintelligenceconferencesin2025 #wheretoattendAIconferencesthisyear