AI Insights on Autonomous Systems with GE's John Lizzi

In the first of our series of A3 interviews with AI leaders, John Lizzi, the Executive Leader - Robotics and Autonomous Systems at GE, discusses how to develop AI projects that focus on business objectives. Lizzi, who serves as the chair of the Association for Advancing Automation’s Artificial Intelligence Technology Strategy Board, says that AI is enabling intelligent systems to operate in the complex and uncertain world. Check out his advice on how to craft your AI strategy.

John Lizzi
Executive Leader - Robotics and Autonomous Systems

How would you advise companies to choose their artificial intelligence projects – and what questions do they need to answer before they begin?

Win hearts and minds: I think it’s important to note that injecting new and disruptive technology into a business is hard no matter what technology you’re talking about. As you choose a project, before you even dive into tech, you need to layout a strategy for winning the hearts and minds of key stakeholders – from the boots on the ground to senior leadership if need be. Each organization is different and therefore strategies will be different. In any case, internal antibodies can kill new tech adoption before it starts if you don’t have hearts and minds.

Mindset: AI is not the goal: At the end of the day, the goal is not to “build AI”, it is to leverage the science and engineering techniques in the field of AI to help deliver a business outcome. You, therefore, need to think of your technology development in the context of building a system to deliver that business outcome – it just happens to leverage a tool in the “AI toolkit” (e.g. NLP (Natural Language Processing), Machine learning, etc.). I’ve seen people get enamored with the magic of AI and end up confused.

Find “dot connectors”: The success stories I’ve seen typically have two common people involved: 1. A person with deep domain/operational understanding and experience, a clear understanding of the value proposition, and an active curiosity about new technology; and 2. Someone with deep experience in building real world AI-enabled systems and who is actively curious about the domain and value. When these 2 “connect the dots”, you have a winning formula. If you don’t have a person #2, hire one or partner with one.

Consider time to value as you select your project – in many cases, you will be under a microscope with a heavy amount of skepticism from those around you on the potential of your AI project. Pick something that’s valuable and that can be built within a reasonable amount of time. The sooner you can prove value, the sooner you’ll garner the hearts and minds of those skeptics, and the sooner you’ll pick up the credibility and momentum to do bigger and bolder things.

Consider data and edge cases early and often:

Data strategy – In my experience, issues with data are a common bottleneck and a frequent cause for technical failure much more often than a failure elsewhere in the system (e.g. the algorithm itself). Do you have the data you need? At the right scale? Of the right quality? Is it accessible? Is your system being fed the most recent and relevant data? Do you have a way for the system to evolve over time as new data and potentially new patterns emerge? If the answer is “no” to any of these questions, you need to step back and reassess, as you may be embarking on a journey to disappointment.

Edge case strategy – in many AI-enabled systems, getting to 90% accuracy is relatively easy. It’s the last 10% that can be extremely difficult, as that 10% is filled with difficult and unexpected situations or “edge cases”. Do you understand your accuracy targets (i.e. is 90% good enough?)? Do you understand your edge cases? What’s your strategy for handling edge cases? Can your system handle them? Do you have a fallback position (e.g. human intervention)?

There’s been a lot of talk about pilot purgatory for AI projects. Companies get a solution running in a lab or a small pilot. But bringing it – at scale – into the real world can be a challenge. How do you overcome this?

A few thoughts:

  1. You can fool a lot of people, and yourself, with demos and proofs of concepts (POC’s). Demos and POC’s can be helpful, but only when they come with a deep understanding of limitations and an understanding of the path to a truly complete and successful solution. Important: The limitations of your system need to be communicated every time a demo is shown – otherwise you may be giving your investors/partners/leadership the latitude to imagine success is much closer than it actually is.
  2. Per my comments above, you need to be able to handle edge cases. In many cases, 95% isn’t good enough. In some cases, 99% isn’t good enough. Consider edge cases early and often and work to address them through iterative POC’s. Humans can be a huge back stop for AI inadequacy if your operations allow for it.
  3. Ensure you’re thinking about the long-term life of your system - In many cases, especially when you’re leveraging machine learning techniques, your “AI” models are trained based upon the data sets and domain assumptions of the time. Those data sets and domain assumptions will change. Your machine learned models also need to change over time accordingly. Ensure you have a strategy for keeping your AI-based system up to date and extending over the life of the system.

What AI application for industry are you most excited about and why?

I’m most excited about advances in AI that are enabling intelligent systems to operate in the complex and uncertain world that we humans live in. My team, and many others, are building uncertainty aware systems that can go beyond the walls of a factory and operate in some very complex and unstructured environments to perform inspections, repairs, field operations, deliveries, etc. The key to making this happen is in building systems that can understand and quantify the variation and uncertainty of the world and adjust to it in human timescales (i.e. as quickly as a human would). The science and engineering techniques of AI, advanced computation, and next generation sensors are helping us do exactly that.

While we all know downtime can help relieve stress, several science-backed studies confirm you actually improve productivity when you take time to enjoy life outside of the office. What are some of your favorite things to do in your free time?

I love spending time with my wife and 4 kids, my parents, and my brother and sister and their families. My family owned an Italian restaurant for almost 40 years, so I love to rekindle that when I get the time. I also love exercising, golf, and an occasional poker game.

John Lizzi is the Executive Leader, Robotics and Autonomous Systems at GE Research where he leads vision, strategy, and technical execution in robotics and autonomous systems. He and his team leverage robotics and AI to deliver value to customers across the DoD, the DoE, GE’s business units, and strategic partners spanning applications in sustainment, field service, avionics, manufacturing, and defense domains. John sits on the Board of Directors of the Association for Advancing Automation (A3), the Technical Advisory Committee of the ARM (Advanced Robotics for Manufacturing) Institute, and chairs the A3’s AI Technology Strategy Board. Over the past 22 years, John has held a number of technical program and leadership roles working closely with GE’s business units and strategic partners. He attended Rensselaer Polytechnic Institute, graduating with an M.S. Degree in Computer and Systems Engineering after graduating with a B.S. Computer Science from Siena College. John also holds an MBA from the State University of New York at Albany. He currently holds 13 patents. John and his wife Kelly have 4 children and reside in Saratoga, NY.