‘AI’ in the sense of machine learning algorithms is very old technology, and has seen revolutionary applications at every step of its development. Eliza was created in the 1960s and demonstrated that if the test of a computer’s artificial intelligence was whether it could fool a human that it was sentient, the value of that test depended very heavily on how willing the human correspondent was to ‘fill in the blanks.’ The results of those experiments show that the average person is extremely willing to fill in the blanks even when the technology is full of gaping holes.
In the podcast, they’re talking specifically about ChatGPT-style technology, its flaws, and the willingness of people to ‘fill in the blanks’ in a new dimension – to assume LLM technology is a truckload more graphics cards or a nuclear reactor away from what sci-fi writers mean when they say ‘Artificial Intelligence’ – and that is evidently false.
‘AI’ in the sense of machine learning algorithms is very old technology, and has seen revolutionary applications at every step of its development. Eliza was created in the 1960s and demonstrated that if the test of a computer’s artificial intelligence was whether it could fool a human that it was sentient, the value of that test depended very heavily on how willing the human correspondent was to ‘fill in the blanks.’ The results of those experiments show that the average person is extremely willing to fill in the blanks even when the technology is full of gaping holes.
In the podcast, they’re talking specifically about ChatGPT-style technology, its flaws, and the willingness of people to ‘fill in the blanks’ in a new dimension – to assume LLM technology is a truckload more graphics cards or a nuclear reactor away from what sci-fi writers mean when they say ‘Artificial Intelligence’ – and that is evidently false.