Put AI to work
June 26-27, 2017: Training
June 27-29, 2017: Tutorials & Conference
New York, NY
 
Grand Ballroom West
11:05am Tackling the limits of deep learning Richard Socher (Salesforce)
11:55am No fluff: Building real, disruptive AI companies Matthew Ocko (Data Collective), Coco Krumme (Haven | UC Berkeley), Friederike Schuur (Fast Forward Labs), Gloria Lau (Unity Medical)
1:45pm Scalable deep learning on AWS using Apache MXNet Guy Ernest (Amazon Web Services)
4:00pm Running TensorFlow at scale in the cloud Yufeng Guo (Google)
4:50pm Benchmarking deep learning inference Sharan Narang (Baidu)
Gramercy East/West
11:55am Transforming an investment firm with AI: A case study Aida Mehonic (The Alan Turing Institute)
1:45pm AI in financial services: Opportunities and obstacles Greg Phalin (McKinsey & Company), Chetan Dube (IPsoft), Doug Kim (Cogito), Aida Mehonic (The Alan Turing Institute)
2:35pm Fighting financial fraud at Danske Bank with artificial intelligence Ron Bodkin (Google), Nadeem Gulzar (Danske Bank Group)
4:00pm Software architectures for building enterprise AI Qirong Ho (Petuum, Inc.)
4:50pm Bayesian deep learning in PyMC3 Thomas Wiecki (Quantopian)
Beekman
11:55am Beyond the hype: Real AI contributions in industry and engineering Christoph Peylo (Bosch Center for Artificial Intelligence)
1:45pm Demystifying AI hype Kathryn Hume (integrate.ai)
2:35pm Is the body the missing link for true AI? Ben Medlock (Microsoft)
4:00pm "Fairness cases" as an accelerant and enabler for AI adoption Chuck Howell (MITRE), Lashon Booker (MITRE)
4:50pm We found a way Tim Estes (Digital Reasoning)
Sutton South/Regent Parlor
11:05am Beyond the state of the art in reading comprehension Jennifer Chu-Carroll (Elemental Cognition)
11:55am Tackling the fake news problem with AI Delip Rao (AI Foundation)
1:45pm What, how, and why: The dynamic of advanced NLG Kristian Hammond (Northwestern Computer Science)
2:35pm Conversational AI at large scale Yishay Carmiel (IntelligentWire)
Murray Hill E/W
1:45pm Building game bots using OpenAI’s Gym and Universe Anmol Jagetia (Media.net)
2:35pm Ray: A distributed execution framework for emerging AI applications Philipp Moritz (University of California, Berkeley), Robert Nishihara (University of California, Berkeley)
4:00pm The biological path toward strong AI Matthew Taylor (Numenta)
4:50pm Building conversational experiences Cathy Pearl (Google)
Sutton Center
11:55am The future of AI is now (sponsored by IBM) Damion Heredia (IBM Watson and Cloud Platform ), Bjorn Austraat (IBM)
4:00pm Top down versus bottom up: Computational creativity Drew Silverstein (Amper Music), Cole Ingraham (Amper Music)
4:50pm Planning for the social impact of AI Madeleine Elish (Data & Society)
Sutton North
11:55am AI: What makes it hard (and fun) (sponsored by Intel) Pradeep Dubey (Intel Corporation)
4:00pm The road to affordable AI-capable products Shaoshan Liu (PerceptIn)
4:50pm Will we automate jobs faster than we create them? Katy George (McKinsey & Company)
Grand Ballroom
9:00am Wednesday opening remarks Ben Lorica (O'Reilly), Roger Chen (Computable)
9:05am Tackling the limits of deep learning Richard Socher (Salesforce)
9:30am Can machines spot diseases faster than expert humans? Suchi Saria (Johns Hopkins University)
9:45am The future of AI is now (sponsored by IBM) Damion Heredia (IBM Watson and Cloud Platform )
9:50am Machines as thought partners David Ferrucci (Elemental Cognition)
10:30am Closing remarks Ben Lorica (O'Reilly), Roger Chen (Computable)
10:35am Morning Break sponsored by IBM | Room: Sponsor Pavilion
3:15pm Afternoon Break Sponsored by Google Cloud | Room: Sponsor Pavilion
5:30pm Sponsor Pavilion Reception | Room: Sponsor Pavilion
7:00pm AI at Night (sponsored by Intel Nervana) | Room: 48 Lounge
12:35pm Lunch sponsored by NVIDIA Wednesday Industry Tables at Lunch | Room: Americas Hall
8:00am Wake up coffee sponsored by Digital Reasoning | Room: 3rd Floor Promenade
8:15am Speed Networking | Room: 3rd Fl Promenade
11:05am-11:45am (40m) Implementing AI Deep Learning
Tackling the limits of deep learning
Richard Socher (Salesforce)
Deep learning has made great progress in a variety of language tasks. However, there are still many practical and theoretical problems and limitations. Richard Socher shares some solutions.
11:55am-12:35pm (40m)
No fluff: Building real, disruptive AI companies
Matthew Ocko (Data Collective), Coco Krumme (Haven | UC Berkeley), Friederike Schuur (Fast Forward Labs), Gloria Lau (Unity Medical)
Join Matt Ocko in conversation with entrepreneurs Hilary Mason, Gloria Lau, and Coco Krumme for a "not your typical" VC panel. They'll discuss how to build disruptive companies that solve real problems with hard AI technologies, digging into the practicalities of getting started, raising money, landing that first big customer, and everything in between.
1:45pm-2:25pm (40m) Implementing AI Cloud, Deep Learning
Scalable deep learning on AWS using Apache MXNet
Guy Ernest (Amazon Web Services)
AWS is democratizing AI, helping you build deep learning systems in any scale, in any team size and skill, and for every use case. Guy Ernest discusses the state of deep learning, the tools that can take advantage of its power, and best practices for building successful businesses in the cloud, including data handling, models learning, deployment, and integration to other parts of the business.
2:35pm-3:15pm (40m) Verticals and applications Machine Learning, User interface and experience, Vision
The science and applications of the emerging field of artificial emotional intelligence
Rana el Kaliouby (Affectiva)
Emotion AI is a branch of artificial intelligence that brings emotional intelligence to AI systems. Rana el Kaliouby reviews the state of emotion AI, its commercial applications, its underlying deep learning methods, and the research roadmap, which includes multimodal emotion recognition and the idea of an emotion chip.
4:00pm-4:40pm (40m) Implementing AI Cloud, Deep Learning
Running TensorFlow at scale in the cloud
Yufeng Guo (Google)
Moving the heavy lifting of machine learning to the cloud is a great way to get large speed-ups. Yufeng Guo walks you through this process in detail so that you'll be ready to scale your own training and prediction services.
4:50pm-5:30pm (40m) Implementing AI
Benchmarking deep learning inference
Sharan Narang (Baidu)
Artificial intelligence has had a tremendous impact on various applications at Baidu, including speech recognition and autonomous driving, although the performance requirements for all of these applications are very different. Sharan Narang outlines the challenges in inference for deep learning models and different workloads and performance requirements for various applications.
11:05am-11:45am (40m) Implementing AI Deep Learning, Financial services
Deep learning applied to consumer transactions with Think Big Analytics
Eric Greene (Think Big Analytics)
Eric Greene compares different approaches to creating models that predict payment amounts, time, and recipient for recurring expenses such as rent, loans, utilities, and services, outlining the data requirements, feature modeling, and neural network architectures that work best, as well as common issues in training and deploying deep learning networks.
11:55am-12:35pm (40m) Verticals and applications Financial services
Transforming an investment firm with AI: A case study
Aida Mehonic (The Alan Turing Institute)
Deploying AI across business functions brings benefits that range from the prosaic to game changers, which in turn also depend on the overall digital and data maturity of the organization. Aida Mehonic shares a case study of an investment firm undergoing an AI transformation across several business units, including trading, reporting, and marketing.
1:45pm-2:25pm (40m)
AI in financial services: Opportunities and obstacles
Greg Phalin (McKinsey & Company), Chetan Dube (IPsoft), Doug Kim (Cogito), Aida Mehonic (The Alan Turing Institute)
AI is changing every area of the financial industry, but the promise of improved performance is accompanied by looming challenges. Greg Phalin leads a panel discussion with Chetan Dube, Doug Kim, and Aida Mehonic on the future of the AI industry, the applicability of AI to use cases in financial services, and the headwinds that could slow adoption of AI at scale.
2:35pm-3:15pm (40m) Verticals and applications Financial services, Machine Learning
Fighting financial fraud at Danske Bank with artificial intelligence
Ron Bodkin (Google), Nadeem Gulzar (Danske Bank Group)
Fraud in banking is an arms race with criminals using machine learning to improve their attack effectiveness. Ron Bodkin and Nadeem Gulzar explore how Danske Bank uses deep learning for better fraud detection, covering model effectiveness, TensorFlow versus boosted decision trees, operational considerations in training and deploying models, and lessons learned along the way.
4:00pm-4:40pm (40m) Implementing AI Hardware, Machine Learning
Software architectures for building enterprise AI
Qirong Ho (Petuum, Inc.)
Petuum, Inc. builds software that lets enterprises develop AI solutions in multiple programming languages and deploy them at scale and with high performance to internal, private computing resources that include a heterogeneous mix of workstations, clusters, CPUs, and GPUs. Qirong Ho outlines the architectural design choices and technical foundation needed to achieve these targets.
4:50pm-5:30pm (40m) Implementing AI Deep Learning, Financial services
Bayesian deep learning in PyMC3
Thomas Wiecki (Quantopian)
Expressing neural networks as a Bayesian model naturally instills uncertainty in its predictions. Thomas Wiecki demonstrates how to embed deep learning in the probabilistic programming framework PyMC3 to address uncertainty and nonstationarity.
11:05am-11:45am (40m) Implementing AI
From ∞ to 8: Translating abstract AI into real numbers for business
Jana Eggers (Nara Logics)
AI has infinite possibilities, but to be adopted by businesses beyond R&D, these solutions must show results. The challenge is that AI often presents new opportunities that aren't easily quantified. Jana Eggers shares lessons learned while taking AI from ideas to results-delivering production solutions at various organizations, including Global 500 enterprises, tech companies, and nonprofits.
11:55am-12:35pm (40m) Implementing AI
Beyond the hype: Real AI contributions in industry and engineering
Christoph Peylo (Bosch Center for Artificial Intelligence)
Generating commercial value from AI in a highly sophisticated industrial environment is a challenge. So far, AI accomplishments in this field stem mostly from marketing rather than systematic application to product lifecycles. Christoph Peylo shares examples of meaningful commercial IoT deployments and discusses obstacles that still have to be overcome.
1:45pm-2:25pm (40m)
Demystifying AI hype
Kathryn Hume (integrate.ai)
Kathryn Hume explores the potential advantages and disadvantages of the AI hype bubble and offers practical tips on how to navigate between real innovation and total nonsense.
2:35pm-3:15pm (40m) Impact of AI on business and society Natural Language, User interface and experience
Is the body the missing link for true AI?
Ben Medlock (Microsoft)
Ben Medlock explores the future of AI, explaining why the potential it holds is not at all frightening. Ben argues that the key to achieving elusive human-like AI lies in a central piece of the puzzle: embodiment.
4:00pm-4:40pm (40m) Impact of AI on business and society Ethics, Governance, and Privacy
"Fairness cases" as an accelerant and enabler for AI adoption
Chuck Howell (MITRE), Lashon Booker (MITRE)
Lack of confidence in the fairness of an AI-based system will limit support for its use and likely preclude adoption, even if that adoption could provide significant benefits. Chuck Howell and Lashon Booker explore tools, techniques, and best practices from the safety-critical software community that can be adapted to provide a “fairness case” framework to address fairness concerns effectively.
4:50pm-5:30pm (40m) Impact of AI on business and society Ethics, Governance, and Privacy, Financial services, Natural Language
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: 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.
11:05am-11:45am (40m) Implementing AI Financial services, Natural Language
Beyond the state of the art in reading comprehension
Jennifer Chu-Carroll (Elemental Cognition)
Why is reading comprehension hard? Jennifer Chu-Carroll offers an overview of current approaches, explaining where they fall short and what our ultimate expectations should be.
11:55am-12:35pm (40m) Media, Natural Language
Tackling the fake news problem with AI
Delip Rao (AI Foundation)
Not a single day goes by without a mention of "fake news" or the problems it causes. Delip Rao offers a nonpartisan overview of fake news, briefly exploring the technology landscape surrounding the content verification and validation problem and diving deeper into the Fake News Challenge and the stance detection problem.
1:45pm-2:25pm (40m) Interacting with AI Media, Natural Language
What, how, and why: The dynamic of advanced NLG
Kristian Hammond (Northwestern Computer Science)
Kristian Hammond offers an overview of advanced natural language generation (NLG), a subfield of artificial intelligence, and the assorted technical systems involved with this emerging technology, along with the mechanisms that drive them.
2:35pm-3:15pm (40m) Interacting with AI Deep Learning, Natural Language, Speech and Voice, User interface and experience
Conversational AI at large scale
Yishay Carmiel (IntelligentWire)
There has been a quantum leap in the performance of conversational AI. From speech recognition to machine translation and language understanding, deep learning made its mark. However, scaling and productizing these breakthroughs remains a big challenge. Yishay Carmiel shares techniques and tips on how to take advantage of large datasets, accelerate training, and create an end-to-end product.
4:00pm-4:40pm (40m) Interacting with AI Deep Learning, Natural Language
Bigger than bots: Machine reading and writing in enterprise
Mohamed Musbah (Maluuba Inc.)
AI research in comprehension, communication, and modeling human-like thinking skills is heralding the dawn of literate machines. Although there has been a lot of recent hype around bots, we’re only just beginning to see the potential for language understanding. Mohamed Musbah explores key research areas and explains how they will power new products and services in language understanding.
4:50pm-5:30pm (40m) Verticals and applications Financial services, Machine Learning, Natural Language
AI-powered natural language understanding applications in the financial industry
Francisco Webber (Cortical.io)
Financial industries are under increased pressure due to regulations that demand extended information management capabilities. Information largely consists of text data, which forces companies to increase headcount to keep up with the growing workload. Francisco Webber demonstrates how Cortical.io’s semantic folding, a neuroscience-based approach to NLU, helps automate these uses cases.
11:05am-11:45am (40m) Implementing AI
AI building AI: How evolutionary algorithms are revolutionizing deep learning
Risto Miikkulainen (Sentient.ai)
Risto Miikkulainen explains how to use massively distributed evolutionary algorithms to evolve the actual architectures of deep networks.
11:55am-12:35pm (40m) Implementing AI Machine Learning
Idea learning: Structuring unstructured data in the enterprise with very little human effort
Ben Vigoda (Gamalon)
Ben Vigoda introduces a new approach to machine learning called idea learning—teaching with ideas instead of labeled data—and demonstrates use cases with state-of-the-art performance in data applications involving structuring of product information, customer feedback, and AI/digital assistant requests.
1:45pm-2:25pm (40m) Interacting with AI Machine Learning, User interface and experience
Building game bots using OpenAI’s Gym and Universe
Anmol Jagetia (Media.net)
Anmol Jagetia explains how to use OpenAI's Gym and Universe to design bots that can become extremely smart using reinforcement learning. You'll create a bot that uses reinforcement learning to beat games and learn how to reuse code to beat a set of games that includes Atari classics (Pac-Man or Pong), a Candy Crush clone, and a racing game.
2:35pm-3:15pm (40m) Implementing AI Machine Learning
Ray: A distributed execution framework for emerging AI applications
Philipp Moritz (University of California, Berkeley), Robert Nishihara (University of California, Berkeley)
AI applications are increasingly dynamic and interactive and work in real time. These properties impose new requirements on the distributed systems that support them. Philipp Moritz and Robert Nishihara offer an overview of Ray, a new system designed to support these emerging applications.
4:00pm-4:40pm (40m) Implementing AI Machine Learning
The biological path toward strong AI
Matthew Taylor (Numenta)
Today's wave of AI technology is still being driven by the ANN neuron pioneered decades ago. Hierarchical temporal memory (HTM) is a realistic biologically constrained model of the pyramidal neuron reflecting today's most recent neocortical research. Matthew Taylor offers an overview of core HTM concepts, including sparse distributed representations, spatial pooling, and temporal memory.
4:50pm-5:30pm (40m) Interacting with AI Cloud, Speech and Voice, User interface and experience
Building conversational experiences
Cathy Pearl (Google)
Brad Abrams explores the latest design and development techniques for building natural language interfaces and draws on the Google Assistant, Actions on Google, and API.AI as examples to explore conversational UI best practices.
11:05am-11:45am (40m)
The practitioner’s guide to AI with Intel Nervana (sponsored by Intel Nervana)
Hanlin Tang (Intel)
Hanlin Tang offers an overview of the Intel Nervana deep learning stack and shares lessons learned from building deep learning solutions for multiple industries.
11:55am-12:35pm (40m) Sponsored
The future of AI is now (sponsored by IBM)
Damion Heredia (IBM Watson and Cloud Platform ), Bjorn Austraat (IBM)
Damion Heredia and Bjorn Austraat explore how augmented intelligence is helping companies disrupt industries and enabling them to make better decisions.
1:45pm-2:25pm (40m)
Embedding machine learning into the fabric of enterprise apps (sponsored by SAP)
Erik Marcade (SAP)
Erik Marcade explains why machine learning and artificial intelligence aren't just revolutionizing industry and knowledge-worker jobs. They're also transforming the way enterprise software is designed and delivered to customers.
2:35pm-3:15pm (40m)
Machine learning with TensorFlow and Google Cloud (sponsored by Google)
Vijay Reddy (Google Cloud)
Vijay Reddy offers a brief overview of TensorFlow, explaining why it's so popular and how to leverage it to build machine learning applications. Vijay walks you through an end-to-end example using TensorFlow for data ingestion, training, and prediction and the Google Cloud Platform to supercharge training and prediction and remove pain from the development and operational workflows.
4:00pm-4:40pm (40m)
Top down versus bottom up: Computational creativity
Drew Silverstein (Amper Music), Cole Ingraham (Amper Music)
Drew Silverstein and Cole Ingraham discuss computational creativity.
4:50pm-5:30pm (40m) Impact of AI on business and society
Planning for the social impact of AI
Madeleine Elish (Data & Society)
When we worry about the Terminator or superintelligence, we miss the social implications of AI that are already beginning to take shape. Madeleine Elish outlines the core challenges to the responsible design and deployment of AI systems and reviews current trends in the ways in which designers and engineers are addressing these challenges across sectors.
11:05am-11:45am (40m)
Accelerating deep learning (sponsored by NVIDIA)
Ryan Olson (NVIDIA)
Ryan Olson explores the role of accelerated GPU computing in modern deep neural networks and explains how it will enable the technologies of the future.
11:55am-12:35pm (40m)
AI: What makes it hard (and fun) (sponsored by Intel)
Pradeep Dubey (Intel Corporation)
We are witnessing a renewed industry interest in machine learning and artificial intelligence and an unprecedented convergence of massive compute with massive data. This confluence has the potential to significantly impact how we do computing and what computing can do for us. Pradeep Dubey shares some of the research Intel is pursuing to enable this compute industry transformation.
1:45pm-2:25pm (40m) Sponsored
Live and let die: The need for an AI-enabled enterprise (sponsored by Arago)
Rene Buest (Arago)
The internet giants are fully embracing AI. The services they offer are all aimed at using data to draw a map of the world, and they are using AI to build disruptive approaches that can't be replicated by established enterprises, which are threatened by these disruptions. However, as Rene Buest explains, most leaders still underestimate the effect this will have on their businesses.
2:35pm-3:15pm (40m)
Bring AI to BI: The benefits of using a GPU database for machine learning and deep learning (sponsored by Kinetica)
Karthik Lalithraj (Kinetica)
Karthik Lalithraj explains how a GPU-accelerated database helps you deploy an easy-to-use, scalable, cost-effective, and future-proof AI solution that enables data science teams to develop, test, and train simulations and algorithms while making them directly available on the same systems used by end users.
4:00pm-4:40pm (40m) Implementing AI Hardware, IoT and its applications
The road to affordable AI-capable products
Shaoshan Liu (PerceptIn)
It is imperative to make high-profile technologies like AI affordable in order for these technologies to proliferate and to benefit the general public. Shaoshan Liu discusses PerceptIn's road to affordable AI-capable products.
4:50pm-5:30pm (40m) Impact of AI on business and society
Will we automate jobs faster than we create them?
Katy George (McKinsey & Company)
The speed with which automation technologies are emerging today and the extent to which they could disrupt the world of work are largely without precedent. How big could the impact be on the world of work, and how rapidly will it be felt? Katy George explores these questions, drawing on a major new report from the McKinsey Global Institute.
9:00am-9:05am (5m)
Wednesday opening remarks
Ben Lorica (O'Reilly), Roger Chen (Computable)
Program chairs Ben Lorica and Roger Chen open the first day of keynotes.
9:05am-9:20am (15m) Deep Learning, Natural Language
Tackling the limits of deep learning
Richard Socher (Salesforce)
AI presents a huge opportunity for businesses to personalize and improve customer experiences and improve efficiency, but the technical complexity of AI puts it out of reach for most companies. Richard Socher explains how Salesforce is doing the heavy lifting to deliver seamless and scalable AI to its customers.
9:20am-9:30am (10m) Sponsored Keynote
AI Now. For Every Industry. (sponsored by NVIDIA)
Jim McHugh (NVIDIA)
AI has the power to transform critical business processes, but new methods will be essential to analyze and visualize data—not as a one-time event but as a continuous process. As a result, a new computing paradigm and deep learning software stack will also be required to power, predict, and act on data to accelerate this transition and produce AI enterprise applications.
9:30am-9:45am (15m)
Can machines spot diseases faster than expert humans?
Suchi Saria (Johns Hopkins University)
Keynote by Suchi Saria
9:45am-9:50am (5m) Sponsored Keynote
The future of AI is now (sponsored by IBM)
Damion Heredia (IBM Watson and Cloud Platform )
Damion Heredia explores how augmented intelligence is helping companies disrupt industries and enabling them to make better decisions.
9:50am-10:05am (15m) Natural Language
Machines as thought partners
David Ferrucci (Elemental Cognition)
AI systems should not only propose solutions or answers but also explain why they make sense. Statistical machine learning is a powerful tool for discovering patterns in data, but, David Ferrucci asks, can it produce understanding or enable humans to justify and take reasoned responsibility for individual outcomes?
10:10am-10:30am (20m) Implementing AI Deep Learning, Machine Learning
Building machines that learn and think like people
Josh Tenenbaum (MIT)
Josh Tenenbaum explains how to build machines that learn and think like people.
10:30am-10:35am (5m)
Closing remarks
Ben Lorica (O'Reilly), Roger Chen (Computable)
Program chairs Ben Lorica and Roger Chen close the first day of keynotes.
10:35am-11:05am (30m)
Break: Morning Break sponsored by IBM
3:15pm-4:00pm (45m)
Break: Afternoon Break Sponsored by Google Cloud
5:30pm-6:30pm (1h)
Sponsor Pavilion Reception
Come enjoy snacks and beverages with fellow O'Reilly AI attendees, speakers, and sponsors.
7:00pm-9:00pm (2h)
AI at Night (sponsored by Intel Nervana)
Join us for an exciting evening filled with delicious food, cocktails, and entertainment at AI at Night, the official attendee party at O'Reilly AI in New York.
12:35pm-1:45pm (1h 10m)
Wednesday Industry Tables at Lunch
Industry Table discussions are a great way to informally network with people in similar industries or interested in the same topics.
8:00am-9:00am (1h)
Break: Wake up coffee sponsored by Digital Reasoning
8:15am-8:45am (30m)
Speed Networking
Gather before keynotes on Wednesday and Thursday morning for a speed networking event. Enjoy casual conversation while meeting new attendees.