Schedule List ViewGrid View

Conference cochairs Ben Lorica, Roger Chen, and Alexis Crowell Helzer open the second day of keynotes.

Keynote by Sean Gourley

Carlos Humberto Morales offers an overview of Nauta, a new open source multiuser platform that allows teams of data scientists to run complex deep learning models on shared hardware resources.

Nick Curcuru, VP, Data Analytics and Cyber Security, will discuss Mastercard’s commitment to AI and its recent investments and developments.

Applied Machine Learning at Facebook

Ruchir Puri discusses the next revolution in automating AI, which strives to deploy AI to automate the task of building, deploying, and managing AI tasks, accelerating enterprises' journey to AI.

Keynote by Christopher Ré

After crossing the first AI implementation milestone, leaders often ask, "What’s next?" Based on experience implementing AI-led automation for more than 100 clients, Accenture has developed an easy-to-use methodology for scaling and sustaining reliable AI solutions. Rajendra Prasad (RP) explains how leaders and change makers in large enterprises can make AI adoption successful.

How can machine learning decode the mysteries of life? Why are algorithms essential to enabling precision medical treatments? How do genomes encode the diversity of cells that make up humans and the signals predisposing us to diseases? Olga Troyanskaya discusses these and other questions through the prism of developing deep learning-based approaches for analysis of the human genome.

The rapid growth of deep learning in demanding large-scale real-world applications has led to a rapid increase in demand for high-performance training and inference solutions. Adam Straw, Adam Procter, and Robert Earhart offer a comprehensive overview of Intel's nGraph deep learning compiler.

Automated machine learning (AutoML) enables both data scientists and domain experts (with limited machine learning training) to be productive and efficient. AutoML is a fundamental shift in how organizations approach machine learning. Francesca Lazzeri and Wee Hyong Tok demonstrate how to use AutoML to automate the selection of machine learning models and automate tuning of hyperparameters.

Arun Kejariwal and Ira Cohen share a novel two-step approach for building more reliable prediction models by integrating anomalies in them. They then walk you through marrying correlation analysis with anomaly detection, discuss how the topics are intertwined, and detail the challenges you may encounter based on production data.

Anoop Katti explores the shortcomings of the existing techniques for understanding 2D documents and offers an overview of the Character Grid (Chargrid), a new processing pipeline pioneered by data scientists at SAP.

Recommender systems support decision making with personalized suggestions and have proven useful in ecommerce, entertainment, and social networks. Sparse data and linear models are a burden, but the application of deep learning sets new boundaries and offers remarkable results. Join Marcel Kurovski to explore a use case for vehicle recommendations at Germany's biggest online vehicle market.

AI planning offers an opportunity to drive reasoning about action trajectories to help build automation. Maja Vukovic demos an application of AI planning for the migration of legacy infrastructure to the cloud, based on real-world examples and data, and discusses challenges in adopting AI planning solutions in the enterprise.

We're entering a new age of software development, where humans and machines work collaboratively together, each doing what they do best. Adam Cheyer offers an overview of a freely downloadable development environment so that you can give this a try yourself and start monetizing your content and services through a new channel that will be backed by more than a billion devices in just a few years.

Although not a universally held goal, maintaining human-centric artificial intelligence is necessary for society’s long-term stability. Joanna Bryson discusses why this is so and explores both the technological and policy mechanisms by which it can be achieved.

There's significant interest in applying deep learning-based solutions to problems in medicine and healthcare. Eric Oermann and Katie Link identify actionable medical problems, recast them as tractable deep learning problems, and discuss techniques to solve them.

Clinical radiology currently faces several clinical issues: improving imaging efficiency, reducing risks, and developing higher imaging quality. Enhao Gong and Greg Zaharchuk explain how Subtle Medical's deep learning/AI solution addresses these problems by enabling faster MRI and faster PET and low-dose scans, providing real clinical and financial benefit to hospitals.

Tammy Bilitzky shares a case study that details lights-out automation and explains how DCL uses AI to transform massive volumes of confidential disparate data into searchable and structured information. Along the way, she outlines considerations for architecting a solution that processes a continuous flow of 5M+ “pages” of complex work units.

Sheldon Fernandez draws on his degrees in both engineering and theology to separate fact from fad in ensuring that artificial systems behave ethically.

While deep learning has shown significant promise for model performance, it can quickly become untenable particularly when data size is short. RNNs can quickly memorize and overfit. Vishal Hawa explains how a combination of RNNs and Bayesian networks (PGM) can improve the sequence modeling behavior of RNNs.

Companies have a strong need for complying with anti-money laundering, antibribery, corruption, and economic sanctions regulation in mitigating third-party risk. Yulia Zvyagelskaya and Victor Llorente highlight how Dow Jones Risk & Compliance uses deep learning and NLP for efficient compliance solutions.

Label leakage is a pervasive problem in predictive modeling data, and it takes on monstrous proportions at enterprise companies, where the data is populated by diverse business processes, making it hard to distinguish cause from effect. Till Bergmann and Leah McGuire explain how Salesforce—which needs to churn out thousands of customer-specific models for any given use case—tackled this problem.

Machine learning models are often susceptible to adversarial deception of their input at test time, which leads to poorer performance. Alina Matyukhina investigates the feasibility of deception in source code attribution techniques in real-world environments and explores attack scenarios on users' identities in open source projects—along with possible protection methods.

Chakri Cherukuri demonstrates how to apply machine learning techniques in quantitative finance, covering use cases involving both structured and alternative datasets. The focus of the talk will be on promoting reproducible research (through Jupyter notebooks and interactive plots) and interpretable models.

Three elements will control the AI market: technology, data, and IP rights. Leveraging rich patent data, Thomas Marlow uncovers the companies with the top patent holdings across the world in groundbreaking research and implementation technologies, surfacing insights into the sources and owners of AI technology as well as the hurdles and opportunities that those entering the field today face.

The significant hype bubble building up around AI has convinced many executives that if they’re not already tech savvy, they might not be ready for AI’s “transformative power.” However, the reality is that AI is just another tool that can help your business, and you’re probably not that far behind. Vinay Seth Mohta explains how to evaluate AI as you would any other strategic investment.

Jana Eggers explores explainability and transparency as both required and unachievable goals for AI, with a focus on helping teams structure discussions about levels of explainability possible and needed for both user trust and regulatory requirements.

Hilary Mason shares a process for repeatedly creating effective AI products, from idea through process to specific design considerations, and explains how architecture and algorithmic choices can support or hinder this process.

Effective data governance is foundational for AI adoption in enterprise, but it's an almost overwhelming topic. Paco Nathan offers an overview of its history, themes, tools, process, standards, and more. Join in to learn what impact machine learning has on data governance and vice versa.

Quantum computers will enable us to efficiently compute things never thought possible, but how will this impact artificial intelligence? Jennifer Fernick explains how to filter signal from noise in discussions surrounding quantum machine learning by exploring how quantum computers work, what types of AI problems they may be good at, and which industries and use cases will (and won't) benefit.

Session with Mohamed Fawzy

Ming-Wei Chang offers an overview of a new language representation model called BERT (Bidirectional Encoder Representations from Transformers). Unlike recent language representation models, BERT is designed to pretrain deep bidirectional representations by jointly conditioning on both left and right context in all layers.

Forough Poursabzi-Sangdeh argues that to understand interpretability, we need to bring humans in the loop and run human-subject experiments. She describes a set of controlled user experiments in which researchers manipulated various design factors in models that are commonly thought to make them more or less interpretable and measured their influence on users’ behavior.

Artificial intelligence has shown great potential to revolutionize clinical medicine and healthcare delivery. However, incorporating these algorithms into clinical workflows involves a big challenge: convincing clinicians and regulators to trust a “black box” solution. Behrooz Hashemian explains how he's helping make deep neural networks interpretable to provide evidence for clinical decisions.

Aric Whitewood details WilmotML's research on the application of AI to investment management and offers an overview of the company's prediction engine, GAIA (the Global AI Allocator), which has been running in production since January 2018.

Andrew Caosun discusses a framework that unifies hidden Markov models and deep learn algorithms (RNN) with modeling components that consider long-term memory and semantics of music (LSTM and convolution). It takes users' original creations as input, modifies the raw scores, and generates musically appropriate melodies.

Diego Oppenheimer draws upon his work with thousands of developers across hundreds of organizations to discuss the tools and processes every business needs to automate model deployment and management so they can optimize model performance, control compute costs, maintain governance, and keep data scientists doing data science.

Building deep learning applications is hard. Building them repeatably is harder. Maintaining high computational performance during a repeatable deep learning development process is borderline impossible. Evan Sparks describes the key pitfalls associated with fast, repeatable model development and details what practitioners can do to avoid them and maintain a supercharged AI development workflow.

Morten Dahl reviews modern cryptographic techniques such as homomorphic encryption and multiparty computation, sharing concrete examples in TensorFlow using the open source library TF Encrypted. Join in to learn how to get started with privacy-preserving techniques today, without needing to master the cryptography.

Magnus Hyttsten explains how to use TensorFlow effectively in a distributed manner using best practices. Magnus covers using TensorFlow's new DistributionStrategies to get easy high-performance training with Keras models (and custom models) on multi-GPU setups as well as multinode training on clusters with accelerators.

Humayun Irshad offers an overview of an active learning framework that uses a crowdsourcing approach to solve parking sign recognition—a real-world problem in transportation and autonomous driving for which a large amount of unlabeled data is available. The solution generates an accurate model, quickly and cost-effectively, despite the unevenness of the data.

ML model and dataset versioning is an essential first step in the direction of establishing a good process. Dmitry Petrov and Ivan Shcheklein explore open source tools for ML models and datasets versioning, from traditional Git to tools like Git-LFS and Git-annex and the ML project-specific tool Data Version Control or DVC.org.

Following the launch of CTRL-labs’s developer kit, CTRL-kit (neural interface device), Patrick Kaifosh paints a picture of a world with neural interfaces, explaining how this technology will change our lives. Patrick outlines a future where we'll be looking up at the world instead of down at our phones—and leads a live demo of CTRL-kit in action.

What's it like to be a mobile phone or to attach a wind sensor to a neural network? Jeff Thompson outlines several recent creative projects that push the tools of AI in new directions. Part technical discussion and part case study for embedding artists in technical institutions, this talk explores the ways that artists and scientists can collaborate to expand the ways that AI can be used.

Adam Marek discusses the motivation for, architecture behind, and functionality of Nauta's offerings and explains how this solution differs from other OSS offerings in the deep learning space.

While deep learning has led to many advancements in computer vision, most research has focused on ground-based imagery. Ryan Mukherjee and Neil Fendley offer an overview of functional Map of the World (fMoW), an ImageNet for satellite imagery built to address this issue, and explain how you can attack or defend these deep learning models.

Jaewon Lee and Sihyeung Han walk you through implementing a self-trained dialogue model using AutoML and the Chatbot Builder Framework. You'll discover the value of AutoML, which allows you to provide better model, and learn how AutoML can be applied in different areas of NLP, not just for chatbots.

Matthew Reyes casts consumer decision making within the framework of random utility and outlines a simplified scenario of optimizing preference on a social network to illustrate the steps in a company’s allocation decision, from learning parameters from data to evaluating the consequences of different marketing allocations.

Ruchir Puri discusses the next revolution in automating AI, which strives to deploy AI to automate the task of building, deploying, and managing AI tasks, accelerating enterprises' journey to AI.

There are many different decisions to make when choosing the right solutions and infrastructure. Drawing on real-world considerations, use cases, and solutions, Nick Curcuru discusses different decisions—and the associated considerations and best practices—Mastercard exercised to build and deploy a successful AI.

Successful financial institutions like Morgan Stanley are growing more committed to efficiency and investing heavily in tools to do so. Marcelo Labre explains how the computing power and AI-readiness of IBM Power Systems enables a new journey of exploration and new possibilities in AI/ML use cases in finance.

Nanda Vijaydev shares practical examples of—and lessons learned from—ML/DL use cases in financial services, healthcare, and other industries. You'll learn how to quickly deploy containerized multinode environments for TensorFlow and other ML/DL tools in a multitenant architecture either on-premises, in the cloud, or in a hybrid environment.

After crossing the first AI implementation milestone, leaders often ask, "What’s next?" Based on experience implementing AI-led automation for more than 100 clients, Accenture has developed an easy-to-use methodology for scaling and sustaining reliable AI solutions. Rajendra Prasad (RP) explains how leaders and change makers in large enterprises can make AI adoption successful.

Many companies want to influence the future by adjusting factors that they control. Suresh Vadakath covers practical ways to extend machine learning models via simulations and points out common pitfalls to avoid. Suresh then discusses a few applications in marketing, pricing, and operations to illustrate how this approach works in the real world.

Topic Table discussions are a great way to informally network with people in similar industries or interested in the same topics.

Ready, set, network! Meet fellow attendees who are looking to connect at the AI Conference. We'll gather before Wednesday and Thursday keynotes for an informal speed networking event. Be sure to bring your business cards—and remember to have fun.