Can ML systems be trusted on their own?

AI systems fail all of the time, which is only a bad thing when a failure is not caught. Failing can be good—it helps us (and AI systems) learn to make better decisions—failing to catch a failure is not. As AI systems automate more work, they have more agency over larger parts of a business, so we need to make sure that their decisions are made accurately. We also should not assume that the decision it makes is always correct. Right now, AI is failing because there’s too little content, it goes out of date quickly, and it doesn’t have the needed level of empathy to be successful—and the maintenance requirement to fix these problems is high, making it a significant barrier to entry for many companies. A team of MIT and IBM Watson researchers is building an efficient, general framework for quantifying how easily deep neural networks can be tricked or misled into making mistakes. Businesses of the future will deploy AI into all facets of their organizations and will need to have a framework where they can transparently understand the risk for error in their machine learning models, see when a system makes a mistake, and know how to fix it. The key is in keeping humans in the loop. In this session, Antony Brydon, CEO at Directly, will discuss how making humans a critical part of AI is key to learning from AI failure.

Who is this presentation for?
Business executives

Prerequisite knowledge:
No

What you'll learn?
Why humans are a critical part of AI learning

Antony Brydon is the CEO and co-founder of Directly, a leader in customer support automation that works with enterprise companies to launch and train virtual agents that double their automation rate. Companies like Airbnb, Microsoft and Samsung use Directly’s expert-in-the-loop AI platform to tap the expertise of their most experienced customers, delivering content, training and answers to their virtual agent to significantly boost performance. Prior to founding Directly in 2011, Antony served as: - Acting CEO for ShopWell, a web and mobile site focused on personal nutrition and food shopping; - Partner at Social Venture Partners, an LLC to invest in and develop internet companies; - CEO and Co-founder of Visible Path, a corporate social networking company funded by Kleiner Perkins and acquired by the Hoovers division of Dun & Bradstreet in January 2008; and - CEO and marketing lead for one of the first music websites, IUMA, which was acquired by EMusic in 1999. Antony graduated from Yale University with a BA in Psychology and Philosophy, and currently resides in San Francisco, CA.