Yole Développement invites you to join us at the Brain Inspired Computing Congress 2020.
The Brain Inspired Computing Congress will bring together the leading start-ups, researchers and multinational companies who are exploring technologies spanning neuromorphic engineering, event-based sensors, brain-inspired algorithms and biologically plausible neural networks. This congress will provide an overview of these technologies in addition to deep-dive sessions on new architectures for neuromorphic chips, event-based sensors, and efforts to create biologically plausible algorithms.
The Brain Inspired Computing Congress will focus on the prime applications for brain-inspired technologies including autonomous vehicles, robotic arm control and dynamic vision sensing. Given the scope for ultra-low-power and edge applications, this technology can be used where conventional deep learning methods are not well suited, such as brain-implants, where it is vital to adhere to power and temperature constraints. Therefore, this congress will also openly discuss the overlap and differentiation between applications for conventional deep learning and brain-inspired computing, exploring how these technologies can complement one another.
Yole Développement will participate in the following:
“Neuromorphic computing, a better solution for the future of Artificial Intelligence“
On April 21-22, 2020
Yohann Tschudi, PhD, Technology & Market Analyst, Computing & Software at Yole Développement
Abstract: Deep Learning is a great technology, some would say revolutionary. But will such superlatives hold true in the future? Looking at where hardware related to these algorithms is going, and what Deep Learning will involve in terms of calculations, we might worry. The neural networks at the heart of these algorithms are getting bigger and demand more computing resources. Today, there are dedicated units within the chips dedicated to calculations typical to neural networks. These neural engines, neural processing units or AI-units are growing in proportion with the neural networks. Another way to run these algorithms is to use Graphical Processing Units (GPUs). These chips fit pretty well, but consume a lot of energy. These two methods, however, are not transferable to ‘edge devices’ such as smartphones, which require low power consumption, small chips. At Yole, we estimate that at current rates, it will be necessary to implement a new type of hardware for AI based on neural networks by five years from now for edge devices. More here.