AI's Cognitive Science Roots Explain Its Unpredictable 'Brain-Like' Behavior
The phrase 'artificial intelligence' emerged during a workshop at Dartmouth College in 1956, proposing that machines could achieve intelligence through established rules. In 1958, Frank Rosenblatt created the Perceptron, which learned from examples and was inspired by the brain, building on Donald Hebb's research regarding neural connections. The 1980s saw David Rumelhart, Geoffrey Hinton, and Ronald Williams advance multilayered neural networks, referred to as 'deep learning.' Contemporary AI learns from examples, resulting in unpredictability and frequent 'hallucinations.' Research by Elizabeth Loftus demonstrated the possibility of implanting false memories. Many users tend to over-rely on AI outputs, viewing them as factual. Educational institutions remain split on AI's educational role, while cognitive science and AI continue to shape one another's theories.
Key facts
- The term 'artificial intelligence' was coined in 1956 at a Dartmouth College workshop.
- Psychologist Frank Rosenblatt built the Perceptron in 1958, an artificial neural network that learned from examples.
- Donald Hebb explored how neural connections strengthen when used, seeding machine learning.
- In the 1980s, David Rumelhart, Geoffrey Hinton, and Ronald Williams developed training for multilayered neural networks, later called 'deep learning.'
- AI 'hallucinations' are common, and thinking of AI as a database makes this behavior confusing.
- Elizabeth Loftus's work showed false memories are common and can be implanted with relative ease.
- A significant proportion of ChatGPT users use it like a search engine, treating outputs as retrieved fact.
- Faculty remain split between banning AI tools outright and building them into coursework.
Entities
Artists
- Frank Rosenblatt
- Donald Hebb
- David Rumelhart
- Geoffrey Hinton
- Ronald Williams
- Elizabeth Loftus
Institutions
- Dartmouth College
- ChatGPT
Locations
- Dartmouth College