-
Members: FreeCIS
IEEE Members: Free
Non-members: FreeLength: 01:17:28
13 Jul 2018
Paul Werbos Keynote Talk (IJCNN) at WCCI 2019
Abstract: Since 2008, there has been a radical growth of AI due to deep learning and the new AI. Google's Tensor Flow initiated a vast influx of commercial and government interest. This was a cultural and technological revolution, not a scientific breakthrough, based on early tools of neural networks and backpropagation. The neural network community has developed much more powerful tools, a well-developed mathematical blueprint for replicating the level of intelligence seen in the mammal brain, and new evidence from brain data that backpropagation is also central to brain intelligence. We are walking a difficult tightrope, as some of us know how to build superintelligent Terminator machines which really would threaten the existence of humans, but as the new mathematics can allow us to understand ourselves better and achieve much higher intelligence and consciousness ourselves. The greatest threat before us here is not from AI as such, but from a rapid haphazard development of the Internet of Things (IOT); a deeper understanding and use of the mathematics of intelligence would be necessary to allow development of a Sustainable human-friendly IOT (SIOT). The mathematics of backpropagation are relevant to machine learning, to Freud's concept of "psychic energy" (where the idea of backpropagation first came from), and to a new vision of what qi and noosphere really are. Www.werbis.com/Mind.htm
Abstract: Since 2008, there has been a radical growth of AI due to deep learning and the new AI. Google's Tensor Flow initiated a vast influx of commercial and government interest. This was a cultural and technological revolution, not a scientific breakthrough, based on early tools of neural networks and backpropagation. The neural network community has developed much more powerful tools, a well-developed mathematical blueprint for replicating the level of intelligence seen in the mammal brain, and new evidence from brain data that backpropagation is also central to brain intelligence. We are walking a difficult tightrope, as some of us know how to build superintelligent Terminator machines which really would threaten the existence of humans, but as the new mathematics can allow us to understand ourselves better and achieve much higher intelligence and consciousness ourselves. The greatest threat before us here is not from AI as such, but from a rapid haphazard development of the Internet of Things (IOT); a deeper understanding and use of the mathematics of intelligence would be necessary to allow development of a Sustainable human-friendly IOT (SIOT). The mathematics of backpropagation are relevant to machine learning, to Freud's concept of "psychic energy" (where the idea of backpropagation first came from), and to a new vision of what qi and noosphere really are. Www.werbis.com/Mind.htm
