Prof. Geoffrey Hinton

A.M. Turing Award

University of Toronto, Canada

  

Learning Deep Neural Networks without Propagating Derivatives

Prof. Abdelaziz Berrado

  

Mohammed V University in Rabat, Morocco

Managing System's Performance in the Age of AI: An Industrial Engineering Perspective

The challenges faced by industrial engineers when dealing with performance issues in complex environments, led to the integration of AI in general and machine learning in particular for addressing societal challenges and improving decision making. In this talk, we illustrate through different applied research projects conducted in Moroccan settings across different industries, businesses, and disciplines, existing opportunities and challenges at the interface between and IE and AI.

Dr. Bouchra Bouqata

  

Microsoft Azure Research, USA

Unleashing the Power of Deep Learning for Robotics Trends and Challenges

The recent rapid progress of deep learning algorithms in generating realistic images, especially in Generative Adversarial Networks (GAN) and Variational Auto-Encoders (VAE), has helped advance new applications. Examples of such applications range from generating and manipulating new synthetic data for self-driving cars, to building/urban architectures, to interior design, and gaming. Furthermore, several applications have benefited from deep learning generative advancement, such as robotics manipulations in structured and unstructured environments, virtual fashion clothes try-on, and item identification on the go. Real-life testing of robotic systems is expensive and slow. A fundamental challenge of applying deep supervised learning and deep reinforcement learning to several applications such as robotics and self-driving cars is data availability and real-world data is difficult to collect. In fact, state-of-the-art deep learning algorithms are data-hungry and need millions to tens of millions of labeled examples with variety in it for each application domain. However, obtaining labeled real-world data for robotics or self-driving cars, is challenging because robots and autonomous cars are expensive, 3-dimensional data can be particularly difficult to label, and collecting data with robotic systems or autonomous vehicles can be dangerous and might not be possible as the underlying algorithms have to reach a certain level of accuracy to be deployed in the wild for data collection. This talk will focus on robotics applications and the use of AI for AI and on how Deep learning algorithms such as GAN are leveraged. Additionally, we will discuss the Machine learning algorithms lifecycle at scale.

Dr. Hicham Zmarrou

Microsoft, Netherlands

To succeed with Enterprise AI, first build the 'Bridge'

Much has been written about the high failure rate of AI projects. AI teams in different organizations often have difficulty moving their insights and algorithms into business processes. At the same time, business teams often can't articulate the problems they need to be solved. And they ignore their cultural resistance to AI innovations. There is a designed-in structural tension between business and AI teams that needs to be recognized and addressed. Structural problems demand structural solutions, and we see a way forward through an AI bridge: an organizational structure and leadership commitment to develop better communication, processes, and trust among all stakeholders. In this presentation we will introduce you to the corporate AI bridge, its role and its major responsibilities.

Prof. Malik Ghallab

LAAS - CNRS, France

  

Ingredients for a National R&D Strategy in AI

During the last decades, AI has moved from the circles of scholarly academic interests to the limelight of media and political concerns, a two sided a fate for a scientific field which brings support as well as risks. Most OECD and many other countries established national plans for AI with significant public and private funding. The establishment of such a plan (which might be under consideration for Morocco) requires analyzing and responding to the following questions:

• Why such a plan is initiated ? what are its intended objectives and how possible progress towards these objectives might be measured ?
• What are the components of the plan? which tasks and actions will it deploy ?
• Who is concerned and what are the actors of the plan ? who will be acting in which task, who will be steering and who will be evaluating ?
• How will the plan be implemented ? how much efforts will be invested and from which sources ? how the tasks and actions are organized and coordinated ?

The talk seeks to clarify the above issues and illustrate possible responses for currently undergoing AI plans in Europe.

Prof. Merouane Debbah

Technology Innovation Institute, UAE

  

Mobile AI: Challenges and Opportunities

Fueled by the availability of more data and computing power, recent breakthroughs in cloud-based machine learning (ML) have transformed every aspect of our lives from face recognition and medical diagnosis to natural language processing. However, classical ML exerts severe demands in terms of energy, memory and computing resources, limiting their adoption for resource constrained edge devices. The new breed of intelligent devices requires a novel paradigm change calling for distributed, low-latency and reliable ML at the wireless network edge. This talk will explore the potential of the Mobile AI paradigm to unlock the full potential of 5G and beyond.

Dr. Priya L. Donti

Climate Change AI, USA

  

Tackling Climate Change with Machine Learning

Climate change is one of the greatest challenges that society faces today, requiring rapid action from all corners. In this talk, I will describe how machine learning can be a potentially powerful tool for addressing climate change, when applied in coordination with policy, engineering, and other areas of action. From energy to agriculture to disaster response, I will describe high impact problems where machine learning can help through avenues such as distilling decision-relevant information, optimizing complex systems, and accelerating scientific experimentation. I will also discuss key considerations for the responsible development and deployment of such work.

Prof. Rachid Guerraoui

EPFL, Switzerland

  

Artificial Intelligence: let's be serious

This talk will discuss the evolution of Artificial Intelligence, highlighting some confusions and open challenges.

Prof. Said Hamdioui

TU Delft - Cognitive IC, Netherlands

  

Computation-in-Memory for Edge AI: Opportunities and Challenges

Evolving IoT applications are extremely demanding in terms of storage, energy efficiency as well as intelligence. Today's computer architectures fail to meet these requirements due to e.g., memory-processor data transfer bottleneck. Hence, alternative computing architectures are being explored in the light of emerging new device technologies.
This talk provides a broad overview of emerging non-volatile devices based computation-in-memory (CIM) architectures for edge AI, gives their potential in enabling smart local energy efficient computing, and highlights the major challenges ahead. The talk first briefly addresses the need of new computing paradigm with energy efficiency of order of fJ/operation to enable zillions of e.g., edge applications, and shows the limitations of both CMOS scaling and today's computing architectures. Then it classifies the sate-of-the art computer architectures and highlight how the trends is going toward computation-in-memory (CIM) architectures in order to eliminated and/or significantly reduces the limitations of today's technologies. The concept of CIM based is discussed and logic and arithmetic circuit designs using such devices and how they enable such architectures are covered. The strong dependency of application domains on the selection of appropriate CIM architecture and its building blocks, as well as the huge potential of CIM (in realizing order of magnitude improvement in terms of energy efficiency) are illustrated based on some case studies. Future CIM challenges including architectures, design, test, and reliability are highlighted.