Tracing the "genetic code" of modern agents back 100 years. Use the timeline to see how the past mirrors the present.
Neurophysiologist Grey Walter built 'Elmer' and 'Elsie', analogue robots that used vacuum tubes to imitate simple neurons. They showed complex behaviour such as phototaxis (seeking light) and avoiding obstacles, and were in effect the first autonomous 'life forms' in robotics.
The core logic of 1949 is still alive today. We just have more computing power.
The first autonomous robots that sought out light.
Neurophysiologist Grey Walter built 'Elmer' and 'Elsie', analogue robots that used vacuum tubes to imitate simple neurons. They showed complex behaviour such as phototaxis (seeking light) and avoiding obstacles, and were in effect the first autonomous 'life forms' in robotics.
Modern equivalent: Modern agentic swarms (AutoGPT)
The birth of neural networks.
Frank Rosenblatt designed the Perceptron, a physical machine that could 'learn' to recognise patterns. It was the ancestor of modern deep learning, using weighted inputs to imitate a biological neuron.
Modern equivalent: Transformers (GPT-4)
The first therapeutic chatbot.
Joseph Weizenbaum created ELIZA, a program that imitated a Rogerian psychotherapist. It showed that people could form emotional bonds with machines based on simple pattern matching and reflection.
Modern equivalent: Claude / ChatGPT
Fixed rules for making decisions.
Systems such as MYCIN used large databases of 'if-then' rules to diagnose diseases. They separated the 'knowledge base' from the 'inference engine', a pattern we see returning today.
Modern equivalent: RAG (Retrieval-Augmented Generation)
Collaborating, autonomous AI systems.
Moving beyond chatbots that answer a single prompt, we are entering the era of 'swarms': specialised agents that work together on complex tasks, much like biological systems such as ant colonies or neural pathways.
Modern equivalent: AISocius ecosystem