(The talk will be given in English)
Speaker: Dr. Daniel Harari
Department of CS and Applied Mathematics, Weizmann Institute
Monday, January 23rd, 2017
15:00 - 16:00
Room 011, Kitot Bldg., Faculty of Engineering
Developing perceptual capabilities in AI systems: Towards human level understanding of the visual world
Abstract
Rapid developments in the field of automated learning have caused a major shift in the approach to the learning of intelligent systems, from explicit instruction to the automatic learning from a large number of labeled examples. Deep neural network models are integrated in the core of new AI technologies such as the autonomous car. Yet, there are still fundamental differences between current AI technologies and human intelligence. In this talk I will present some examples of these differences, including computational models that demonstrate how an AI system can learn complex visual concepts (such as hand recognition, gaze estimation and spatial relations) rapidly and without explicit supervision like humans, and in particular infants, do.
Short Bio
Research interests:
My research interests lie in the interdisciplinary area between computer and human vision, including perception, cognition and developmental learning. The goal is to develop human level understanding of the visual world in AI systems, using computational models with cognitive capabilities.
Education:
- A senior intern at the computer science department working with Prof. Shimon Ullman (2016-now)
- Postdoc at MIT in the Center for Brains, Minds and Machines (CBMM) (2013-2015)
- PhD in computer science from the Weizmann institute (2012)
- MSc and BSc in electrical engineering (MSc - Tel Aviv University, 2000; BSc – Technion, 1994)
Industrial experience:
- Advisor to the image processing team at El-Op (2010-2013)
- Video content analysis researcher at Nice Systems (2000-2007)
- Digital signal processing engineer at IDF (1994-2000)
quantum cascade lasers (THz-QCLs) founded on designed electron dynamics in less than handful of subbands. As a demonstrative case study for this capability, I will review my efforts to untangle the complexity of the temperature driven electron transport in THz-QCLs towards extending their operation to room temperature (
). This study has led to the experimental demonstration of negative differential resistance at room temperature in THz-QCL structures. This is a strong evidence for the possibility to achieve lasing at room temperature. The study points also to a future direction in quantum engineering that is to control and manipulate generated heat and carriers’ temperature for the benefit of the device and for the realization of novel device concepts.
