Presented By O’Reilly and Intel AI
Put AI to Work
April 29-30, 2018: Training
April 30-May 2, 2018: Tutorials & Conference
New York, NY
 
Grand Ballroom East
11:05am
1:45pm
Add Deep Sentiment Analysis Across Language Boundaries to your personal schedule
4:00pm Deep Sentiment Analysis Across Language Boundaries Gerard de Melo (Rutgers University)
4:50pm
Grand Ballroom West
Add Learned Index Structures to your personal schedule
11:55am Learned Index Structures Tim Kraska (MIT)
Add Revolutionising Aviation with AI  to your personal schedule
1:45pm Revolutionising Aviation with AI Carolina Sanchez Hernandez (NATS)
Add Classify Images in Spark to your personal schedule
2:35pm Classify Images in Spark Dave Nielsen (Intel)
Add The Search for a New Visual Search, Beyond Language to your personal schedule
4:00pm The Search for a New Visual Search, Beyond Language Mike Ranzinger (Shutterstock)
Sutton North/Center
11:05am
1:45pm
4:00pm
Sutton South
Add The Vital Role of Failure in Machine Learning to your personal schedule
11:55am The Vital Role of Failure in Machine Learning Scott Weller (SessionM)
1:45pm
Add Best Practices for Machine Learning in the Enterprise to your personal schedule
2:35pm Best Practices for Machine Learning in the Enterprise Robbie  Allen  (InfiniaML)
Regent Parlor
1:45pm
Add Collaborative machine intelligence: Accelerating human knowledge  to your personal schedule
4:00pm Collaborative machine intelligence: Accelerating human knowledge Emily Pavlini (Diffeo), Max Kleiman-Weiner (Diffeo)
Nassau East/West
Add Racial Bias in Facial Recognition Software to your personal schedule
11:55am Racial Bias in Facial Recognition Software Stephanie Kim (Algorithmia)
Add Model Evaluation in the Deep Learning Land to your personal schedule
4:00pm Model Evaluation in the Deep Learning Land Pramit Choudhary (DataScience.com)
4:50pm
Concourse A
Add Determining Normal (and Abnormal) Using Deep Learning to your personal schedule
11:05am Determining Normal (and Abnormal) Using Deep Learning John Hebeler (Lockheed Martin)
Add Fooling Neural Networks in the Physical World to your personal schedule
11:55am Fooling Neural Networks in the Physical World Andrew Ilyas (Massachusetts Institute of Technology), Logan Engstrom (Massachusetts Institute of Technology), Anish Athalye (Massachusetts Institute of Technology)
Add How to Save Time Optimizing Chatbots to your personal schedule
2:35pm How to Save Time Optimizing Chatbots Ofer Ronen (Chatbase (via Area 120; Area 120 is an incubator for early-stage products operated by Google))
Add Do-it-yourself Artificial Intelligence to your personal schedule
4:00pm Do-it-yourself Artificial Intelligence Alasdair Allan (Babilim Light Industries)
Add Online and Active Learning for Recommender Systems  to your personal schedule
4:50pm Online and Active Learning for Recommender Systems Jorge Silva (SAS Institute Inc)
8:55am Keynotes to come
Add Wednesday Topic Tables at lunch to your personal schedule
12:35pm Lunch Wednesday Topic Tables at lunch | Room: America's Hall
Add Speed Networking to your personal schedule
8:00am Speed Networking | Room: TBD
10:35am Morning Break | Room: Sponsor Pavilion
3:15pm Afternoon Break | Room: Sponsor Pavilion
11:05am-11:45am (40m)
Session
11:55am-12:35pm (40m) Implementing AI, Models and Methods
End to end video analytics solution to surveillance and secure high value assets
Harsh Kumar (Intel Corp)
Using AI and Computer vision for security surveillance in the energy industry
1:45pm-2:25pm (40m) Models and Methods
Session
2:35pm-3:15pm (40m) Implementing AI
From answering questions to questioning answers: Challenges of large scale QnA systems
Mridu Narang (Microsoft)
Knowledge acquisition techniques for users in a world of information overload and information manipulation are expected to provide instant, precise and succinct answers. Question Answering Systems are faced with the challenges of serving answers with high accuracy and backed by strong verification techniques. This talk offers an overview of challenges & approaches of such large scale QnA systems.
4:00pm-4:40pm (40m) Models and Methods
Deep Sentiment Analysis Across Language Boundaries
Gerard de Melo (Rutgers University)
Across the globe, people are voicing their opinion online. However, sentiment analysis is challenging for many of the world's languages, particularly with limited training data. This talk shows how one can instead exploit large amounts of surrogate data to learn advanced word representations that are custom-tailored for sentiment, and presents a special deep neural architecture to use them.
4:50pm-5:30pm (40m)
Session
11:05am-11:45am (40m) Implementing AI
Deep Reinforcement Learning’s Killer App: Intelligent Control in Real-World Systems
Mark Hammond (Bonsai)
Reinforcement learning is a powerful machine learning technique for solving problems in dynamic and adaptive environments. Mark Hammond dives into two real-world case studies and demonstrates how to build and deploy deep reinforcement learning models for industrial applications.
11:55am-12:35pm (40m) Implementing AI, Interacting with AI, Models and Methods
Learned Index Structures
Tim Kraska (MIT)
I will present how fundamental data structures can be enhanced using machine learning with wide reaching implications even beyond indexes. To quote Steven Sinofsky, board partner at A16Z and former president at Microsoft: "This paper [about learned indexes] blew my mind....ML meets 1960's data structures and crushes them."
1:45pm-2:25pm (40m) Implementing AI
Revolutionising Aviation with AI
Carolina Sanchez Hernandez (NATS)
The Aviation industry is awakening to new technologies that can revolutionise the way it operates and evolve. Airports in particular are perfect candidates for AI and Machine Learning concepts. NATS is collaborating with several companies and institutes to change the way that data is captured and processed to transform Airport operations.
2:35pm-3:15pm (40m) Implementing AI, Models and Methods
Classify Images in Spark
Dave Nielsen (Intel)
BigDL enables Deep Learning frameworks natively for Apache Spark. Using BigDL, we created a new app to demonstrates how you can use image recognition. To demonstrated Deep Learning in Apache Spark, we developed an app called VegNonVeg. The app uses BigDL framework to classify images of food as vegitarian or non-vegitarian.
4:00pm-4:40pm (40m) Models and Methods
The Search for a New Visual Search, Beyond Language
Mike Ranzinger (Shutterstock)
Mike Ranzinger of Shutterstock will detail his research on composition aware search. He will demonstrate how the research led to the launch of AI technology allowing users to more precisely find the image they need within Shutterstock’s collection of more than 150 million images. The goal of this research was to improve the future of search using visual data, contextual search functions, and AI.
4:50pm-5:30pm (40m) Implementing AI
Reliable and Robust Classification Pipeline for Protein Crystallisation Imaging
Christopher Watkins (CSIRO)
The achievement of human level accuracy in image classification through the use of modern AI algorithms has renewed interest its application to automated protein crystallisation imaging. This session aims to discuss the development of the deep tech pipeline required for the robust operation of an on-line classification system in CSIRO's GPU cluster and the lessons learned along the way.
11:05am-11:45am (40m)
Session
11:55am-12:35pm (40m) Implementing AI, Models and Methods
Scaling up Deep Learning Based Super Resolution Models more efficiently using Cloud
Xiaoyong Zhu (Microsoft)
In this talk, we will demonstrate the latest academic progress in Super Resolution field using Deep Learning, especially GANs, and how it can be applied in various industries, such as Medical area; we will also showcase how the training could be done in a distributed fashion on the Cloud, which gives more torch light to researchers and data scientists who are in this empirical field.
1:45pm-2:25pm (40m) Implementing AI
Session
2:35pm-3:15pm (40m) Implementing AI
An Open Extensible AI Platform implementing 4 Use Cases for the Enterprise
Murali Kaundinya (Merck)
An innersource model to curate and operationalize Machine Learning and Deep Learning algorithms with a common workflow and engaging user experience. With a focus on patterns and practices, this talk shares experiences realizing 4 enterprise scale use cases namely optical character recognition, release engineering, virtual customer assistants and data unification,
4:00pm-4:40pm (40m)
Session
4:50pm-5:30pm (40m) Implementing AI
The long and winding road to AI: Lessons from implementing Cognitive AI
Rupert Steffner (WUNDER.ai)
The way to real-world AI is a long and winding road. All what we heard from reputable experts turned out to be true: The need for better data, a new UX, new ways of learning and many more. This session highlights the lessons we have learnt while implementing cognitive AI applications to help consumers finding the products they love. With evidence what to expect in case you build AI too lean.
11:05am-11:45am (40m) AI Business Summit, AI in the Enterprise
Using Artificial Intelligence to Enhance the Digital Experience
Ron Bodkin (Google)
AI innovation in Deep Learning has moved from labs to large-scale deployments at Google. Hear techniques and lessons from Google, Spotify, Netflix, Evernote and eBay. Learn how to apply AI for personalized recommendations, intelligent bot support and enhancing the digital experience. Hear how to overcome common pitfalls in dealing with data, automation, experimentation.
11:55am-12:35pm (40m) AI Business Summit, AI in the Enterprise
The Vital Role of Failure in Machine Learning
Scott Weller (SessionM)
In video games, players learn by failing. They might “die” hundreds of times before learning how to succeed. By enabling us to simulate scenarios and predict outcomes, AI has essentially made the world like a game that we can play with, yet we expect immediate success. Is this realistic? Technologist Scott Weller explores the role of failure in machine learning using real-world examples.
1:45pm-2:25pm (40m) AI Business Summit, AI in the Enterprise
Session
2:35pm-3:15pm (40m) AI Business Summit, AI in the Enterprise
Best Practices for Machine Learning in the Enterprise
Robbie  Allen  (InfiniaML)
Learn from an entrepreneur behind two successful AI companies who implemented machine learning and NLP solutions in over a hundred organizations. Find out the factors common to successful machine learning implementations and which factors predict failure. Finally, you will learn how to build and cultivate ML talent within your organization in an increasingly competitive job market.
4:00pm-4:40pm (40m) AI Business Summit, AI in the Enterprise
Lessons Learned Through Building an AI Company from the Ground Up
Nicole Eagan (Darktrace)
Although AI technology seems to be everywhere, implementing AI in practice is a real challenge. The technology needs to be scalable, trusted by the humans that use them, and easily accessible for those with limited AI expertise. With over 4 years’ experience and 4,000 deployments, Darktrace has unique insights into how to develop and deploy both practical and successful AI applications.
4:50pm-5:30pm (40m) AI Business Summit, Impact of AI on Business and Society
Democracy, Human Rights and Rule of Law by design for Artificial Intelligence
Paul Nemitz (European Commission)
The advent of Artifical Intelligence requires a new innovation model. With the learning, judging and deciding machine becoming ever more pervasive, it is necessary to insert by design and default the basic rules of democracy, human rights and the rule of law into the innovation process and the programmes of Articial Intelligence. The presentation gives examples on how this can be done.
11:05am-11:45am (40m) AI Business Summit, AI in the Enterprise, Impact of AI on Business and Society
Executive Briefing: Beyond Killer Robots - A Framework for the Ethical Implementation of AI
Emma Kinnucan (Booz Allen Hamilton)
Executives responsible for leading AI are dedicating too much effort dispelling myths about implications of the technology (e.g., “killer robots” and mass unemployment). More realistic ethical issues concerning AI’s impact on human privacy, equity, dignity, and justice are ignored.We’ll provide explicit, real-world examples that show ignoring AI’s immediate ethical implications is a business risk.
11:55am-12:35pm (40m) AI Business Summit, AI in the Enterprise
Executive Briefing: Building a learning organization is AI's hat trick
Jana Eggers (Nara Logics)
AI scores points for providing better answers to your company's challenges. It also gets points for requiring you to get your data house in order. I think AI's hat trick is how it can transform your company into a learning organization. I'll review the benefits of a learning org, as well as the key aspects, then show you how to build an AI program that can support you in achieving those benefits.
1:45pm-2:25pm (40m) AI Business Summit, AI in the Enterprise, Impact of AI on Business and Society
Session
2:35pm-3:15pm (40m) AI Business Summit, AI in the Enterprise
Executive Briefing: Lean AI Product Development (and common pitfalls)
Shane Lewin (Lumiata)
Great AI products are more than technology; they are built on a clear (computationally tractable) model of customer success. Getting that model right can more challenging than building the AI models themselves; and getting it wrong is very expensive. In this talk we cover common pitfalls in defining AI products, and organizing teams to solve them; and talk through emerging best practices.
4:00pm-4:40pm (40m) AI Business Summit, Interacting with AI, Models and Methods
Collaborative machine intelligence: Accelerating human knowledge
Emily Pavlini (Diffeo), Max Kleiman-Weiner (Diffeo)
Recent advances have made machines more autonomous. However, much remains to be done in order for AI collaborate with people. Drawing inspiration from the way humans accumulate knowledge and naturally work together, we will share new insights that enable machines and people to work and learn as a team, discovering new knowledge in unstructured natural language content together.
4:50pm-5:30pm (40m) AI Business Summit, Implementing AI, Interacting with AI
From Here to "Her": Evolving Chatbot Interactions to Meet the Relational Needs of Humans
Ian Beaver (Next IT), Cynthia Freeman (Next IT)
As conversation emerges as the next great human-machine interface, we discuss the challenges faced by the AI industry to relate to humans in the way they relate to each other. Highlighting findings from a recent study we demonstrate relational strategies used by humans in conversation and how Virtual Assistants must evolve to communicate effectively.
11:05am-11:45am (40m) Models and Methods
Avoiding Biased Algorithms: Lessons from the Hiring Space
Lindsey Zuloaga (HireVue)
We are all familiar with the highly-publicized stories of algorithms displaying overtly biased behavior towards certain groups. What is actually happening behind the scenes and how can these situations be avoided? In this session, Lindsey Zuloaga (HireVue) will share experiences and lessons learned in the hiring space to help others avoid unfair modeling and work to establish best practices.
11:55am-12:35pm (40m) Implementing AI
Racial Bias in Facial Recognition Software
Stephanie Kim (Algorithmia)
This talk will cover the basics of facial recognition and the importance of having diverse datasets when building out a model. We’ll explore racial bias in datasets using real world examples and cover a use case for developing an OpenFace model for a celebrity look-a-like app.
1:45pm-2:25pm (40m) Models and Methods
Combining well-established statistical techniques with modern machine learning algorithms
Funda Gunes (SAS Institute)
As machine learning algorithms and artificial intelligence continue to progress, we must take advantage of the best techniques from various disciplines. In this presentation, I will show how combining well-proven methods from classical statistics can enhance modern deep learning methods in terms of both predictive performance and interpretability.
2:35pm-3:15pm (40m) Models and Methods
Recurrent Neural Networks for Recommendations and Personalization
Nick Pentreath (IBM)
In the last few years, RNNs have achieved significant success in modeling time-series and sequence data, in particular within the speech, language and text domains. Recently, these techniques have been begun to be applied to session-based recommendation tasks, with very promising results. In this talk I explore the newest research advances in this domain, as well as practical applications.
4:00pm-4:40pm (40m) Implementing AI, Interacting with AI
Model Evaluation in the Deep Learning Land
Pramit Choudhary (DataScience.com)
Just predicting the target label for computer vision machine learning problem is not enough. It is also important to understand the “why”, “what” and “how” about the categorization process. In this talk we explore different ways to faithfully interpret and evaluate Deep Neural Network models - CNN Image models to understand the impact of salient features in driving categorization.
4:50pm-5:30pm (40m)
Session
11:05am-11:45am (40m) Models and Methods
Determining Normal (and Abnormal) Using Deep Learning
John Hebeler (Lockheed Martin)
Determining abnormal conditions depends on maintaining a useful definition of normal. Deep learning methods can track dynamic, complex operations to create a normal operating envelop for dynamic, data-rich environments. Self-organizing maps dynamically group activities while recurrent neural networks predict their likelihood for residency in an identified group.
11:55am-12:35pm (40m) Interacting with AI, Models and Methods
Fooling Neural Networks in the Physical World
Andrew Ilyas (Massachusetts Institute of Technology), Logan Engstrom (Massachusetts Institute of Technology), Anish Athalye (Massachusetts Institute of Technology)
We’ve developed an approach to generate 3D adversarial objects that reliably fool neural networks in the real world, no matter how the objects are looked at.
1:45pm-2:25pm (40m) Implementing AI, Interacting with AI, Models and Methods
TensorFlow Lite: how to accelerate your Android and iOS app with AI
Kazunori Sato (Google)
TensorFlow Lite is TensorFlow’s lightweight solution for Android, iOS and embedded devices. It enables on-device machine learning inference with low latency and a small binary size. In this session, we will discuss how developers can use TensorFlow Lite to overcome the challenges for bringing the latest AI technology to production mobile apps and embeded systems.
2:35pm-3:15pm (40m) AI in the Enterprise
How to Save Time Optimizing Chatbots
Ofer Ronen (Chatbase (via Area 120; Area 120 is an incubator for early-stage products operated by Google))
Chatbots are expected to make machine communication feel human. But high-quality bot experiences are very hard to build. Certain issues in particular make building bots that do not frustrate users difficult. This presentation will delve into such issues and suggest ways, including machine learning, for developers to save time addressing them.
4:00pm-4:40pm (40m) Implementing AI
Do-it-yourself Artificial Intelligence
Alasdair Allan (Babilim Light Industries)
The AIY Projects kits bring Google's machine learning algorithms to developers with limited experience in the field, allowing them to prototype machine learning applications and smart hardware more easily. We walkthrough how to setup and build the kits, and how to use the kits Python SDK to use machine learning both in the cloud, and locally on the Raspberry Pi.
4:50pm-5:30pm (40m) Implementing AI, Models and Methods
Online and Active Learning for Recommender Systems
Jorge Silva (SAS Institute Inc)
Recommender systems suffer from concept drift and scarcity of informative ratings - we use a Bayesian approach to tackle both problems by making the learning process online and active. Online learning deals with concept drift, i.e., changing user preferences. Active learning prioritizes the most informative users and items by quantifying uncertainty in a principled, probabilistic framework.
8:55am-10:35am (1h 40m)
Plenary: Keynotes to come
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.
8:00am-8:30am (30m)
Speed Networking
Ready, set, network! Meet fellow attendees who are looking to connect at The AI Conference. We'll gather before Wednesday keynotes to host an informal speed networking event. Be sure to bring your business cards and have fun.
10:35am-11:05am (30m)
Break: Morning Break
3:15pm-4:00pm (45m)
Break: Afternoon Break