Meet Our Speakers

Meet Our Sponsors


Speaker: Lewis Shepherd – Sr. Technology Advisor, Department of Defense

Speaker: John Shegerian, Co-founder and Executive Chairman, ERI


Protecting digital data in the emerging world of AI is a terrain where the our nation’s most significant entities are missing a crucial piece of the puzzle!

Government agencies, defense contractors, and prominent, category-leading corporations are leaders in pioneering A.I. technology, but they are too often overlooking a key component…protecting digital data from hardware hacking.

With A.I. innovation comes tremendous capacity for management of digital data – from trade secrets to homeland security protocols. And while protection of data in the realm of cybersecurity continues to improve, the area that gets overlooked – and the potential cause of devastating breaches – is the protection of data stored on the A.I. devices themselves. How susceptible are these devices to breaches? What happens to them at the end of life or when components are upgrade and replaced?  Where does all the data go?

In this presentation, John Shegerian, co-founder and Executive Chairman of ERI, the nation’s largest recycler of electronic waste and the world’s largest cybersecurity-focused hardware destruction company, will share areas of vulnerability in A.I. (and other high-tech platforms) and what can be done to shore up effective security for data protection in a rapidly evolving A.I. universe.

Panelists: Rand Waltzman – Deputy CTO RAND Corporation; Osonde Osoba, Ph.D – Professor, Pardee RAND Graduate School, Information Scientist; Elsa Kania, Adjunct Fellow, CNAS

Abstract: AI as a field has seen many ups and downs in terms of public interest and support since the first AI conference in 1956 where the term Artificial Intelligence was coined.  Periods of euphoria were followed by downturns that came to be known as “AI Winters.”  The panel discussion will take up such questions as (1) Are we seeing the beginning of the next AI Winter? (2) What might the next AI Winter look like? (3) What steps can we as a community take to minimize the risks and possible consequences of another AI Winter? Realistic expectations are critical to successfully sustaining a program whose goal is to apply any new technology.  This is a particularly sensitive issue for technologies like AI that are the subject of grossly exaggerated and non-stop coverage in the media.  

Speaker: John Sipple – Machine Learning scientist at Google and the Defense Innovation Unit at the Department of Defense)

Today, penetration testing is performed manually with tools like Nmap, Metasploit, the Social Engineering Toolkit. These tools provide the security professional both the sensors and actions to perform the task of discovering and exploiting vulnerabilities. We will review some key fundamental concepts of function approximation, dimensionality reduction, and describe how deep neural networks map sensors to actions, and how reinforcement learning is effective at choosing novel and complex actions and maximizing long-term reward. Applying these fundamentals, we present an open source architecture that connects Deep Reinforcement Learning with some of the common pen-testing tools that can be used for automatically discovering and exploiting network vulnerabilities.

Speaker: Ryan Lewis, VP, Deputy Director, In-Q-Tel CosmiQ Works

The last few years has seen a significant increase in the launch of commercial and federal remote sensing platforms ranging from small satellites to unmanned aerial systems (UAS). The proliferation of these systems increased both the volume and access to geospatial data especially imagery and full motion video. As a result, traditional and nontraditional end users alike now have unparalleled access robust and diverse data streams. At the same time,advanced analytic techniques like machine learning, specifically computer vision, have advanced rapidly with a heavy reliance on open source algorithms, associated tools, and software frameworks. The combination of these two technology trends promises to change how insight is extracted from geospatial data. Understanding how end users worked with these data and insights will be critical. This presentation will discuss a proposed framework for defining levels of analyst augmentation that will allow mapping contributors and machine learning researchers to better understand each other and help direct the application of these advanced algorithms against mapping problems.

Applying AI techniques to Human Languages can enhance the productivity of knowledge workers. However, this potential can be impeded by challenges related to language representation, data quality, and scalability.


Speaker: Kfir Bar (Chief Scientist, Basis Technolgy)
Abstract: To achieve the promises of AI, algorithms in the field of Natural Language Processing (NLP) need to comprehend events and named entities in text with human or beyond-human accuracy. In this talk, we will survey recent results of applying AI techniques on some NLP tasks, such as named entity recognition and event extraction. We will discuss the difference we see in accuracy and speed, while comparing some of the best known deep architectures with a basic statistical approach. We will discuss the challenges of processing languages in multiple languages, such as English, Arabic and Korean.
The main focus of this talk is to discuss the practical, industry-level, considerations for deploying deep cutting-edge AI solutions, including prediction speed, accuracy and explainability.