Jürgen Schmidhuber

∙

0 followers

Jürgen Schmidhuber is an informatician renowned for his work on artificial intelligence, deep learning and artificial neural networks. He is co-director of the Dalle Molle Institute in Manno, Lugano district in Ticino, South of Switzerland, for Artificial Intelligence Research. He is sometimes called the ‘father of AI’ or ‘father of deep learning.’ Schmidhuber studied at the Munich Technical University in Germany. He was a professor of artificial intelligence at the Università della Svizzera Italiana in Lugano, Switzerland from 2004 until 2009.

Schmidhuber and Sepp Hochreiter published a paper in 1997 on a type of recurrent neural network, called Long-Term Memory (LSTM) in English. In 2015, LSTM was used in the new implementation of Google Smartphone software speech recognition. Google also used LSTM for the Allo and Google Translate smart assistants. For the “Quicktype” function on the iPhone and for Siri, Apple used LSTM. Amazon used Amazon Alexa’s LSTM. In 2017, Facebook carried out some 4.5 billion automatic translations on LSTM networks each day. “Theses powers make LSTM the likely commercial AI achievement used for all things, from disease prediction to the composition of music,” said Bloomberg Business Week. “In the 2011 IDSIA team Schmidhuber and his postdoc Dan Ciresan also achieved dramatic speedups of convolutionary neural networks on fast-parallel computers called GPUs. An earlier CNN of Chellapilla et al. GPU was 4 times faster than the CPU equivalent. Dan Ciresan et al.’s deep CNN at IDSIA was 60 times faster and reached its first superhuman performance in a computer vision contest in August 2011. Their quick, deep CNNs won no less than four picture competitions between May 15, 2011 and September 10, 2012. They also improved significantly on the best literature performance for several image databases. The approach is now central to computer vision. It is based on CNN designs introduced by Yann LeCun et al. many years earlier and which have applied the backpropagation algorithm to a variant of the CNN architecture of Kunihiko Fukushima, which was later modified by J. In 2014, Schmidhuber established a Nnaisense company to work on artificial intelligence business applications in areas such as finance, heavy industry and self-drive cars. The company is advised by Sepp Hochreiter, Jaan Tallinn and Marcus Hutter. Sales were below USD 11 million in 2016; Schmidhuber notes, however, that research and not revenue are the current focus. In January 2017, Nnaisense raised its first round of capital funding. The overall objective of Schmidhuber is to create an entirely useful AI through sequence training of a single AI in a range of tight tasks; however skeptics point out that companies like Arago GmbH and IBM have been using AI for various projects for years, without showing any signs of artificial general intelligence.

The Guardian reports that Schmidhuber complained in a “scathing 2015 article” that fellow students Geoffrey Hinton, Yann LeCun, and Yoshua Bengio are “huge quotations” each other but “not lent credit to the pioneers,” allegedly underlining Schmidhuber and other pioneers of early machine learning, including Alexey Grigorevich Ivakhnenko, who first published the article. The Guardian reports that

In 2013, Schmidhuber was awarded the Helmholtz Award of the International Neural Network Society and in 2016, the IEEE Computational Intelligence Society’s Neural Networks Pioneer Award. He is a member of the European Academy of Arts and Sciences.