Jürgen Schmidhuber is a German computer scientist widely regarded as a founding father of modern deep learning and recurrent neural networks (RNNs). He is the Director of the Swiss AI lab IDSIA and a professor of computer science at KAUST.
The LSTM Network
In 1997, Schmidhuber and his student Sepp Hochreiter co-invented the Long Short-Term Memory (LSTM) recurrent neural network. By addressing the "vanishing gradient problem" that made standard RNNs unable to remember long sequences, LSTM enabled neural networks to process speech, handwriting, and text sequences. By 2015, LSTM was run on billions of smartphones, powering speech recognition for Google, Apple (Siri), and translation systems.
AGI and Metalearning
Schmidhuber's research also spans metalearning (learning to learn), unsupervised learning, reinforcement learning, and formal theories of artificial general intelligence (AGI), including the Gödel Machine, which is a mathematically rigorous self-referential self-improving AI system.
