BREAKING
ELIZA: The 1966 Chatbot Illusion
0
build started
0
first described
0
met PARRY
How ELIZA Faked a Conversation
1
scan keywords
↓
2
decompose input
↓
3
reassemble reply
↓
4
echo as question
Rule-Based vs Neural AI
ELIZA (1966)
symbolic
●
MAD-SLIP on IBM 7094
●
No learning, no knowledge
●
Pattern matching only
Modern LLMs
neural
●
Trained on vast data
●
Fluent, capable output
●
Same emotional pull
The ELIZA Effect
Persuasion Isn't Intelligence
AI NEWS BLITZ
Nearly six decades before ChatGPT, an MIT program revealed how easily we trust machines.