Presented By O’Reilly and Intel Nervana
Put AI to work
September 17-18, 2017: Training
September 18-20, 2017: Tutorials & Conference
San Francisco, CA
 
Grand Ballroom
11:05am No session
1:45pm No session
Add Applications of neural-based models for conversational speech to your personal schedule
2:35pm Applications of neural-based models for conversational speech Yishay Carmiel (IntelligentWire)
Add Using deep learning toolkits with Kubernetes clusters to your personal schedule
4:00pm Using deep learning toolkits with Kubernetes clusters Wee Hyong Tok (Microsoft), Joy Qiao (Microsoft)
Yosemite A
Add AI for manufacturing: Today and tomorrow to your personal schedule
11:55am AI for manufacturing: Today and tomorrow David Rogers (Sight Machine)
Add Deep learning on mobile: The how-to guide to your personal schedule
1:45pm Deep learning on mobile: The how-to guide Anirudh Koul (Microsoft)
2:35pm No session
4:00pm No session
Yosemite BC
Add Conversational AI in Amazon Alexa to your personal schedule
11:55am Conversational AI in Amazon Alexa Ashwin Ram (Amazon)
Add Active learning and transfer learning to your personal schedule
1:45pm Active learning and transfer learning Lukas Biewald (CrowdFlower)
Add Reinforcement Learning Overview  to your personal schedule
4:00pm Reinforcement Learning Overview Melanie Warrick (Google)
Imperial A
Add Self-supervised visual learning and synthesis to your personal schedule
11:05am Self-supervised visual learning and synthesis Alyosha Efros (UC Berkeley)
Add Adversarial machine learning to your personal schedule
11:55am Adversarial machine learning Alex Kurakin (Google)
Add End-to-end deep learning at the edge  to your personal schedule
1:45pm End-to-end deep learning at the edge Bruno Fernandez-Ruiz (Nexar)
Add All the linear algebra you need for AI to your personal schedule
2:35pm All the linear algebra you need for AI Rachel Thomas (fast.ai)
Add Scalable deep learning to your personal schedule
4:00pm Scalable deep learning Ameet Talwalkar (Determined AI)
Imperial B
Add When machines have ideas: A new approach to AI to your personal schedule
11:05am When machines have ideas: A new approach to AI Ben Vigoda (Gamalon)
Add Evolving neural networks through neuroevolution to your personal schedule
11:55am Evolving neural networks through neuroevolution Kenneth Stanley (Uber AI Labs | University of Central Florida)
Add Why AI needs human-centered design to your personal schedule
4:50pm Why AI needs human-centered design James Guszcza (Deloitte Consulting)
Franciscan AB
Add It's the organization, stupid. to your personal schedule
11:05am It's the organization, stupid. Jana Eggers (Nara Logics)
Add We found a way. to your personal schedule
11:55am We found a way. Tim Estes (Digital Reasoning)
Franciscan CD
Add Intel Xeon scalable processor architecture and AI workload performance to your personal schedule
1:45pm Intel Xeon scalable processor architecture and AI workload performance Banu Nagasundaram (Intel), Akhilesh Kumar (Intel)
Add The practitioner’s guide to AI to your personal schedule
2:35pm The practitioner’s guide to AI Hanlin Tang (Intel)
Add Wednesday opening remarks to your personal schedule
Grand Ballroom B
8:50am Wednesday opening remarks Ben Lorica (O'Reilly Media), Roger Chen
Add How to escape saddle points efficiently to your personal schedule
8:55am How to escape saddle points efficiently Michael Jordan (UC Berkeley)
Add Accelerating AI to your personal schedule
9:35am Accelerating AI Steve Jurvetson (DFJ)
Add Fireside chat with Naveen Rao and Steve Jurvetson to your personal schedule
9:50am Fireside chat with Naveen Rao and Steve Jurvetson Naveen Rao (Intel), Steve Jurvetson (DFJ)
Add Our Skynet moment to your personal schedule
10:05am Our Skynet moment Tim O'Reilly (O'Reilly Media)
Add Closing remarks to your personal schedule
10:25am Closing remarks
Add Wednesday Topic Tables at lunch to your personal schedule
12:35pm Lunch sponsored by Microsoft Wednesday Topic Tables at lunch | Room: Grand Ballroom B
10:35am Morning Break sponsored by Google Cloud | Room: Sponsor Pavilion
3:15pm Afternoon Break | Room: Sponsor Pavilion
Add Speed Networking to your personal schedule
8:10am Speed Networking | Room: Yosemite Foyer
11:05am-12:35pm (1h 30m)
Session: No session
1:45pm-2:25pm (40m)
Session: No session
2:35pm-3:15pm (40m) Implementing AI Deep learning
Applications of neural-based models for conversational speech
Yishay Carmiel (IntelligentWire)
Today almost every achievement in language understanding is based on neural networks. Yishay Carmiel explains why analyzing conversational speech is still a challenging proposition despite all the recent breakthroughs in natural language processing and offers some potential solutions.
4:00pm-4:40pm (40m) Implementing AI Data science and AI, Deep learning
Using deep learning toolkits with Kubernetes clusters
Wee Hyong Tok (Microsoft), Joy Qiao (Microsoft)
Joy Qiao and Wee Hyong Tok demonstrate how to combine Kubernetes clusters and deep learning toolkits to get the best of both worlds and jumpstart the development of innovative deep learning applications. Along the way, Joy and Wee Hyong explain how to train deep neural networks using GPU-enabled containers orchestrated by Kubernetes with common deep learning toolkits, such as CNTK and TensorFlow.
4:50pm-5:30pm (40m) Impact on business and society Data science and AI, Deep learning
Personalizing ecommerce for two billion people on Facebook
Nikita Lytkin (Facebook)
Nikita Lytkin explains how Facebook uses machine learning technologies developed by its ads ranking, applied machine learning, and AI research teams to enable personalized ecommerce that recommends a vast diversity of products to nearly two billion people.
11:05am-11:45am (40m) Impact on business and society Security, Technical best practices
Incident response evolved: How AI is revolutionizing how we combat cyberthreats
Aaron Goldstein (Cylance)
The current threat landscape is in a state of evolution that poses a significant risk to organizations' assets, reputations, and identities. Aaron Goldstein explores new and existing threats (and why traditional defenses fail to address them) and explains how leveraging AI techniques can improve the speed and efficiency of incident response tactics, even when combating the toughest threat actors.
11:55am-12:35pm (40m) Verticals and applications Data science and AI, IoT (including smart cities, manufacturing, smart homes/buildings)
AI for manufacturing: Today and tomorrow
David Rogers (Sight Machine)
Artificial intelligence in manufacturing has been around for a long time, but are you aware of how it can make your operations more efficient and profitable? David Rogers explains how existing technologies like the digital twin approach, advanced decision making, and downtime cause detection have primed manufacturing for a profitable and efficient future.
1:45pm-2:25pm (40m) Implementing AI Deep learning, Tools and frameworks
Deep learning on mobile: The how-to guide
Anirudh Koul (Microsoft)
Over the last few years, convolutional neural networks (CNN) have risen in popularity, especially in computer vision. Anirudh Koul explains how to bring the power of deep learning to memory- and power-constrained devices like smartphones.
2:35pm-3:15pm (40m)
Session: No session
4:00pm-5:30pm (1h 30m)
Session: No session
11:05am-11:45am (40m)
We're in the final stretch for—and early innings of—autonomous vehicles: Fireside chat with Shahin Farshchi and Ashu Rege
Shahin Farshchi (Lux Capital), Ashu Rege (Zoox)
Join Shahin Farshchi in conversation with Ashu Rege, who is reinventing the automobile from scratch at Zoox to offer consumers an unforgettable, autonomous transportation experience. They'll discuss early challenges that turned out to be straightforward, easy problems that turned out to be very hard, and the obstacles that lie ahead.
11:55am-12:35pm (40m)
Conversational AI in Amazon Alexa
Ashwin Ram (Amazon)
Conversational AI in Amazon Alexa
1:45pm-2:25pm (40m) Implementing AI Technical best practices, Transportation and autonomous vehicles
Active learning and transfer learning
Lukas Biewald (CrowdFlower)
Making the best possible use of training data is essential for effective machine learning. Active learning can make your training data collection 10x–1,000x more efficient, while transfer learning opens up a world of new training data possibilities. Lukas Biewald explores the state of the art in training data, active learning, and transfer learning, especially as applied to deep learning.
2:35pm-3:15pm (40m) Implementing AI Algorithms, Enterprise adoption
Deep reinforcement learning in the enterprise: Bridging the gap from games to industry
Mark Hammond (Bonsai)
Mark Hammond explores how enterprises can move beyond games and leverage deep reinforcement learning and simulation-based training to build programmable, adaptive, and trusted AI models for their real-world applications.
4:00pm-4:40pm (40m) Implementing AI Algorithms, Data science and AI
Reinforcement Learning Overview
Melanie Warrick (Google)
Reinforcement learning is a popular subfield in machine learning because of its success in beating humans at complex games like Go and Atari. The field’s value is in utilizing an award system to develop models and find more optimal ways to solve complex, real-world problems. This approach allows software to adapt to its environment without full knowledge of what the results should look like.
4:50pm-5:30pm (40m) Implementing AI Data and training, Deep learning
TensorFlow, machine learning, and learning to learn
Sherry Moore (Google)
TensorFlow is the world's most popular machine learning framework. Google Brain team member Sherry Moore discusses the latest developments in TensorFlow and offers a dive into her research on evolving deep learning models using genetic algorithms.
11:05am-11:45am (40m)
Self-supervised visual learning and synthesis
Alyosha Efros (UC Berkeley)
Alyosha Efros shares several case studies exploring the paradigm of self-supervised learning and discusses several ways of defining objective functions in high-dimensional spaces. Alyosha also covers the applications of this technology for image synthesis, including automatic colorization, image-to-image translation, curiosity-based exploration, and, terrifyingly, #edges2cats.
11:55am-12:35pm (40m)
Adversarial machine learning
Alex Kurakin (Google)
Adversarial machine learning session by Alex Kurakin
1:45pm-2:25pm (40m) Implementing AI Data science and AI, Deep learning, IoT (including smart cities, manufacturing, smart homes/buildings), Transportation and autonomous vehicles
End-to-end deep learning at the edge
Bruno Fernandez-Ruiz (Nexar)
Current driving policy models are limited to models trained using homogenous data from a small number of vehicles running in controlled environments. Bruno Fernandez-Ruiz offers an overview of a network of connected devices that is building an end-to-end driving policy to leverage the 10 trillion miles driven every year.
2:35pm-3:15pm (40m) Implementing AI Deep learning, Tools and frameworks
All the linear algebra you need for AI
Rachel Thomas (fast.ai)
If the math used in AI seems intimidating, this tutorial is for you. Rachel Thomas walks you through working with arrays of different dimensions and how broadcasting handles data dimensions. You'll also gain hands-on experience with PyTorch, the Python framework for GPU computing developed by Facebook.
4:00pm-4:40pm (40m)
Scalable deep learning
Ameet Talwalkar (Determined AI)
Ameet Talwalkar offers an overview of Hyperband, a novel algorithm for hyperparameter optimization that is simple, flexible, theoretically sound, and an order of magnitude faster than leading competitors, and shares research aimed at understanding the underlying landscape of training deep learning models in parallel and distributed environments.
4:50pm-5:30pm (40m)
Industrial robotics and deep reinforcement learning
Derik Pridmore (Osaro)
There continues to be a gap between the most advanced papers and the reality of deployed industrial robots. Derik Pridmore explores the most recent advances in deep and reinforcement learning for robotics, the current state of industrial robotics, and how Osaro is working to bridge the gap.
11:05am-11:45am (40m) Implementing AI Algorithms, Data and training
When machines have ideas: A new approach to AI
Ben Vigoda (Gamalon)
Ben Vigoda demonstrates new advances in AI technology that enable companies to accurately read millions of complex customer messages and take action.
11:55am-12:35pm (40m) Implementing AI Algorithms, Deep learning, Tools and frameworks, Transportation and autonomous vehicles
Evolving neural networks through neuroevolution
Kenneth Stanley (Uber AI Labs | University of Central Florida)
Kenneth Stanley offers an overview of the field of neuroevolution, an emerging paradigm for training neural networks through evolutionary principles that has grown up alongside more conventional deep learning, highlighting major algorithms such as NEAT, HyperNEAT, and novelty search, the field's emerging synergies with deep learning, and promising application areas.
1:45pm-2:25pm (40m)
High-performance computing opportunities in deep learning
Greg Diamos (Baidu)
Accuracy scales with data and compute, transforming some difficult AI problems into problems of computational scale. Greg Diamos covers challenges to further improving performance and outlines a plan of attack for tearing down the remaining obstacles standing in the way of strong scaling deep learning to the largest machines in the world.
2:35pm-3:15pm (40m) Implementing AI Architectures, Deep learning
Why do we need new hardware for machine intelligence?
Nigel Toon (Graphcore)
Nigel Toon explains how new processing platforms will enable the next wave of machine intelligence beyond deep learning and how these machine learning innovations will impact businesses and improve competitiveness.
4:00pm-4:40pm (40m) Implementing AI New product development, Transportation and autonomous vehicles
Enabling computer-vision-based autonomous driving with affordable and reliable sensors
Shaoshan Liu (PerceptIn)
Autonomous cars, like humans, need good eyes and a good brain to drive safely. Shaoshan Liu explains how PerceptIn designed and implemented its high-definition, stereo 360-degree camera sensors targeted for computer-vision-based autonomous driving.
4:50pm-5:30pm (40m)
Why AI needs human-centered design
James Guszcza (Deloitte Consulting)
James Guszcza shares the principles of human-computer collaboration, organizes them into a framework, and offers several real-life examples in which human-computer cognitive collaboration has been crucial to the economic success of a project.
11:05am-11:45am (40m) Impact on business and society Case studies, Enterprise adoption
It's the organization, stupid.
Jana Eggers (Nara Logics)
Having spent the last three years working with Global 200 customers to get AI systems into production, Jana Eggers can tell you that the technology is (finally) ready—but the organization is not. Jana discusses the top five reasons orgs struggle—data silos, the tech-business gap, driving innovation, resistance to change, and the hype-reality gap—and shares ideas on how to overcome them.
11:55am-12:35pm (40m) Impact on business and society Data science and AI, Deep learning
We found a way.
Tim Estes (Digital Reasoning)
As AI moves from concept to reality, debates about ethics are evolving into excitement and the desire to learn more about AI and its promise of a better world. Tim Estes discusses two customer use cases at Digital Reasoning: Nasdaq, which found a way to use AI to help safeguard financial markets, and Thorn, which found a way to use AI to combat human trafficking and rescue children.
1:45pm-2:25pm (40m) Impact on business and society Bots, Deep learning
Critical factors in building successful AI-powered conversational interfaces
Paul Tepper (Nuance Communications)
Many industries are now exploring chatbots powered by artificial intelligence as a source for improved insights and better understanding of customer preferences. Paul Tepper explores the unique challenges chatbots present, shares available solutions, and outlines a number of critical factors in building successful chatbots and virtual assistants.
2:35pm-3:15pm (40m) Implementing AI Law, ethics and governance (including AI safety), Tools and frameworks
Ensuring smarter-than-human intelligence has a positive outcome.
Nate Soares (MIRI)
The field of artificial intelligence has made major strides in recent years, but there is a growing movement to consider the implications of machines that can rival humans in general problem-solving abilities. Nate Soares outlines the underresearched fundamental technical obstacles to building AI that can reliably learn to be "aligned" with human values.
4:00pm-4:40pm (40m) Verticals and applications Algorithms, Enterprise adoption
Let artificial intelligence handle one of the two certainties in this world
Gang Wang (Intuit)
Taxes are one of consumers' most complex financial transactions, thanks to a tax code that is 80,000 pages long. Gang Wang explains how Intuit built the industry’s only Tax Knowledge Engine, a constraint-based engine that encodes changing financial regulations and provides the foundation for a host of artificial intelligence technologies that save customers time and money.
4:50pm-5:30pm (40m) Verticals and applications Finance, Healthcare
Vertical AI: Solving full stack industry problems using subject-matter expertise, unique data, and AI to deliver a product's core value proposition
Bradford Cross (DCVC)
Low-level task-based AI gets commoditized quickly, and more general AI is decades off. While most of the machine learning talent works in big tech companies, massive, timely problems lurk in every major industry outside tech. Bradford Cross explains how vertical AI startups leverage subject-matter expertise, AI, and unique data to deliver their product's core value proposition.
11:05am-11:45am (40m) Sponsored
Scalable deep learning with Microsoft Cognitive Toolkit (sponsored by Microsoft)
Anusua Trivedi (Microsoft)
Anusua Trivedi offers an overview of Microsoft’s Cognitive Toolkit, also known as CNTK. CNTK has unique advantages over other toolkits, especially in speed and scalability. Anusua compares five well-known toolkits to demonstrate how CNTK achieves almost linear scalability, which is far superior to all the other well-known toolkits.
11:55am-12:35pm (40m) Sponsored
Pushing the boundaries of ML using TensorFlow and Google Cloud (sponsored by Google Cloud)
Magnus Hyttsten (Google)
Magnus Hyttsten explains how Google is pushing the boundaries of machine learning with TensorFlow and Google Cloud, sharing some of the latest models Google teams have been working on and the technical challenges they've encountered, new APIs in TensorFlow, how the Tensor Processing Unit (TPU) works, and how Google Cloud can be used to train extremely large models.
1:45pm-2:25pm (40m)
Intel Xeon scalable processor architecture and AI workload performance
Banu Nagasundaram (Intel), Akhilesh Kumar (Intel)
Banu Nagasundaram and Akhilesh Kumar offer an overview of the architectural features of the latest Intel Xeon scalable processor, outline the changes from previous generations, and discuss the architectural benefits that favor AI workloads. Along the way, Banu and Akhilesh explore AI workload performance for data center CPUs.
2:35pm-3:15pm (40m) Implementing AI
The practitioner’s guide to AI
Hanlin Tang (Intel)
Training deep learning networks is often seen as a dark art. Hanlin Tang demystifies the process, sharing lessons learned from building AI algorithms across multiple verticals and tips and tricks for designing models. Hanlin also offers an overview of the Intel Nervana deep learning stack, which accelerates the iteration cycle for data scientists.
8:50am-8:55am (5m)
Wednesday opening remarks
Ben Lorica (O'Reilly Media), Roger Chen (.)
Program chairs Ben Lorica and Roger Chen open the second day of keynotes.
8:55am-9:10am (15m)
How to escape saddle points efficiently
Michael Jordan (UC Berkeley)
Many new theoretical challenges have arisen in the area of gradient-based optimization for large-scale data analysis, driven by the needs of applications and the opportunities provided by new hardware and software platforms. Michael Jordan shares recent research on the avoidance of saddle points in high-dimensional nonconvex optimization.
9:10am-9:25am (15m)
Why democratizing AI matters: Computing, data, algorithms, and talent
Jia Li (Google)
Jia Li has contributed to some of the most influential datasets in the world and helped transform computer vision from an academic niche into a dominant technology. Jia explains why a democratized approach to AI ensures that the compute, data, algorithms, and talent behind these technologies reach the widest possible audience.
9:25am-9:35am (10m) Sponsored Keynote
AI mimicking nature: Flying and talking (sponsored by Microsoft)
Lili Cheng (Microsoft)
Lili Cheng shares two examples of AI inspired by nature. In the first, Microsoft researchers created a system that uses artificial intelligence that draws on the way birds fly to keep a sailplane aloft. The second explores what makes people unique, our language instinct, and our ability to model how people socialize and accomplish work.
9:35am-9:50am (15m)
Accelerating AI
Steve Jurvetson (DFJ)
Keynote by Steve Jurvetson
9:50am-10:00am (10m)
Fireside chat with Naveen Rao and Steve Jurvetson
Naveen Rao (Intel), Steve Jurvetson (DFJ)
Join Naveen Rao and Steve Jurvetson for a fireside chat.
10:00am-10:05am (5m) Sponsored Keynote
Build smart applications with your new super power: Cloud AI (sponsored by Google Cloud)
Philippe Poutonnet (Google)
Google has invested deeply in machine learning for many years and is using it successfully across its highly successful consumer businesses. Philippe Poutonnet explains how to leverage the power of ML with Google Cloud, using the platform's powerful data management tools, support for collaborative experiments, and predictions at Google scale.
10:05am-10:25am (20m)
Our Skynet moment
Tim O'Reilly (O'Reilly Media)
Tim O’Reilly draws on lessons from networked platforms to show how our economy and financial markets have also become increasingly managed by algorithms, making the case that income inequality, declining upward mobility, and job losses due to technology are not inevitable; they are the result of design choices we have made in the algorithms that manage our markets.
10:25am-10:35am (10m)
Closing remarks
Program chairs Ben Lorica and Roger Chen close the second day of keynotes.
12:35pm-1:45pm (1h 10m)
Wednesday Topic Tables at lunch
Topic Table discussions are a great way to informally network with people in similar industries or interested in the same topics.
10:35am-11:05am (30m)
Break: Morning Break sponsored by Google Cloud
3:15pm-4:00pm (45m)
Break: Afternoon Break
8:10am-8:40am (30m)
Speed Networking
Gather before keynotes on Tuesday and Wednesday morning for a speed networking event. Enjoy casual conversation while meeting new attendees.