Short answers, pulled from the story.
Rina Dechter introduced the term deep learning to the machine learning community in 1986. The field focuses on utilizing multilayered neural networks to perform tasks such as classification and regression.
The Group method of data handling was the first working deep learning algorithm published by Alexey Ivakhnenko and Lapa in 1965. They regarded it as a form of polynomial regression generalizing Rosenblatt's perceptron to handle complex relationships.
Yann LeCun created a CNN called LeNet in 1989 for recognizing handwritten ZIP codes on mail requiring three days of training. LeNet-5 released in 1998 classified digits and was applied by several banks to recognize numbers on checks digitized in 32x32 pixel images.
Researchers at Google DeepMind and Lawrence Berkeley National Laboratory announced GNoME discovered over 2 million new materials within a short timeframe in November 2023. The system achieved a noteworthy success rate of 71 percent validated through autonomous robotic experiments.
Goertzel hypothesized these behaviors stem from limitations in internal representations inhibiting integration into heterogeneous multi-component artificial general intelligence architectures. Some deep learning architectures display problematic behaviors such as confidently classifying unrecognizable images as belonging to familiar categories in 2014.