Job Title: Director

Affiliation:  Silicon Valley AI Lab, Baidu Research

Bio: Adam Coates is Director of the Silicon Valley AI Lab at Baidu Research in Sunnyvale. He received his PhD in 2012 from Stanford University and subsequently was a post-doctoral researcher at Stanford. During his graduate career, Coates co-developed the most sophisticated autonomous helicopter in the world, capable of advanced aerobatic maneuvers. He also worked on perception systems for household robots before studying deep learning algorithms. His thesis work investigated issues in the development of deep learning methods, particularly the success of large neural networks trained from large datasets. He also led the development of large scale deep learning methods using distributed clusters and GPUs. At Stanford, his team trained artificial neural networks with over 11 billion connections—the largest in the world—using techniques for high performance computing systems

Talk Title: Deep Speech (Video , Slides)

Abstract: Speech recognition technology promises to dramatically alter our relationship with machines, making devices and the internet easier to use and more accessible. However,  current systems are not sufficiently accurate or natural for users to rely on them.  This is especially critical in developing economies where new users will connect to the internet for the first time with a mobile device.  As part of our mission to create technologies that let us have an impact on hundreds of millions of people, our team in the Silicon Valley AI Lab at Baidu is creating a new type of speech recognition system, called DeepSpeech, that seeks to significantly improve speech recognition performance and eventually make devices as natural to interact with as other humans.  In this talk, I will describe how scalability and Deep Learning are driving progress in AI, enabling powerful end-to-end systems like DeepSpeech to reach new levels of performance.

Job Title: Chief Technology Officer

Affiliation: AlgoSnap Inc.

Bio: Dr. Emmanuel Munguia Tapia received his MS and PhD degrees from the Massachusetts Institute of Technology (MIT) and has over 15 years of multi-disciplinary expertise combining machine learning, context awareness and novel sensor hardware to make mobile, wearable and IoT devices smarter. He is presently the cofounder and chief technology officer of AlgoSnap Inc., a startup focused on creating intelligent algorithms and analytics from massive sensor datasets to achieve unprecedented IoT device intelligence. Dr. Munguia Tapia was previously a technologist and strategist at Intel Corporation in the areas of machine learning, context awareness, sensors, and biosensors. As the former director of Context Awareness and Mobile Sensing for Samsung he won the Samsung Gold Medal Award for creating the most innovative technology company wide in 2014. He was also the recipient of the 10 year impact award at UBICOMP 2014, the top International Joint Conference on Pervasive and Ubiquitous Computing.

Dr. Munguia Tapia also holds a diploma in Engineering Leadership from University of California Berkeley and has 36+ international publications and 10+ patents.

Talk Title: The Magic Box: Automating the Development Cycle for Intelligent Algorithms (Video , Slides)

Abstract: Traditionally, the development of robust intelligent agents and algorithms for commercial products is a slow and inefficient process that takes months or even years. We estimate that up to 70% of the resources in industry and even academia are spent on tasks that do not advance the state-of-the-art for intelligent algorithms. Instead, an enormous amount of time is spent building data collection infrastructures, managing data collection campaigns, experimenting with existing machine learning techniques, and finally porting the algorithms from analytics tools (e.g., Matlab) to target devices where they are tested and optimized. In this talk, we will review these pain points and present our platform which accelerates the algorithm development cycle over 100x while making it less expensive using automation, crowdsourcing, and crowdfunding so that businesses can ship algorithms in days instead of years. We will also introduce CrowdSignals.io, an effort to ethically collect the largest mobile dataset across the United States so progress on intelligent algorithms that use sensor data can catch-up with progress made in other areas such as vision via large datasets such as ImageNet and enable the application of techniques such as Deep learning.

Abstract:  In a world where Virtual Agents are gradually demonstrating increasing levels of intelligence, users still find virtual agents lacking in human-like capabilities and qualities. One of the key factors in this gap in human-likeness is Multidimensional Intelligence. Many developers and are still focusing on different levels of functional intelligence, while leaving aside other types of intelligence that have a decisive impact on the users’ perception. As we will see, it is all about perception: when intelligence cannot be furthered within the current state of the art, the appearance of intelligence can still boost the user’s overall experience. Cognitive Design and a solid persona in conjunction with Social and Emotional Intelligence can effectively help bridge the gap that brings agents closer to their ultimate goal.

Panel discussion with: (Video)

 - Ron Kaplan, A9

 - Peter Yeh, Nuance

 - Adam Coates, Baidu

 - Emmanuel Munguia Tapia, AlgoSnap

 - Pilar Manchón, Intel

 - Moderator: Bo Begole, Huawei