AlianzaFuturista
This video also contains the Q&A session with David Wood, Ben Goertzel and Sophia. / Este vídeo también contiene la sesión de preguntas a David Wood, Ben Goertzel y Sophia.
TransVision 2018 Madrid
Ateneo de Madrid
19, 20 y 21 de octubre, 2018
Promotores: José Luis Cordeiro, Humanity Plus, Life Extension Foundation
Organizadores: José Luis Cordeiro, Breyssi Arana, Alba Montero
Vídeo: Sergio M.L. Tarrero para Alianza Futurista
Cámaras: Iñigo Cores, Sergio M.L. Tarrero, Pablo Tarrero
Ayudantes de producción: David García Monaj, Paco Mota
ENGLISH:
TransVision 2018 Madrid
Ateneo de Madrid
October 19, 20 & 21, 2018
Promoters: José Luis Cordeiro, Humanity Plus, Life Extension Foundation
Organizers: José Luis Cordeiro, Breyssi Arana, Alba Montero
Video: Sergio M.L. Tarrero for Alianza Futurista
Cameras: Iñigo Cores, Sergio M.L. Tarrero, Pablo Tarrero
Production Assistants: David García Monaj, Paco Mota
Source
Hawkins of Numenta got his money from phase 1. Now he has a lifestyle company researching human and machine intelligence.
All the research in associative memory that was done in the 1960s has been forgotten. However it is the means of giving deep neural networks working, short, medium and long term memory. Thus making them far more powerful.
If you have a locality sensitive hash with +1,-1 binary outputs you can weight and sum those outputs to give a recalled value. To train first you recall and calculate the error. Divide by the number of bits. Add or subtract that number as appropriate to each weight to make the error zero. Through the magic of the central limit theorem you have an associative memory system.
For the LSH you can use predetermined random sign flipping followed by the fast Walsh Hadamard transform to give a random projection. Then binarize to get the LSH.
Basically you can expect to see jumps in the performance of deep neural networks as new (or old) ideas are combined in. Even with no further technical innovations they will have a major impact.