On Monday I had the pleasure of speaking on an Advertising Week panel alongside Julia Huang, MD at AMEX Ventures, and Jaya Kolhatkar, CDO at Hulu. The talk, moderated by Forbes reporter Marty Swant, discussed what it means to implement AI into a business strategy from the perspective of people who build or manage AI for some of the biggest companies in the world. The three of us shared our thoughts on data, AI, and where we're heading. While our backgrounds and experiences vary, we were in agreement on a few key points throughout the discussion:
1. It’s all about the data.
Everyone is collecting data in some form, but it’s how you use that data that is most important. Julia said AMEX is always trying to test the effectiveness of data. And Jaya said that Hulu collects data on every second of use — every search, every show that’s watched, every time those shows are paused, and so forth. In all, that adds up to approximately 15 billion interactions every day. IV.AI manage similar data for all of our customers from internal Customer data to external inputs from social, websites, and articles.
But it’s crucial to remember that not all data is good data. Quality controls built around testing can help ensure that the data you receive is ultimately useful in the long run. Jaya pointed out that it's crucial to put in the time to ensure you have a good foundation and can trust your data. And as I've learned, even unsupervised learning is somewhat supervised because you are choosing the data that goes in, which can open up your work to bias.
And just remember: It’s better to use less data that is of higher quality than to use more data that is questionable. Data needs to be strong and lead to a straightforward experience that empowers the consumer.
2. Identify the problems you want AI to solve.
Successful AI implementations boil down to solving a central issue that can’t otherwise be solved.
