The Importance of Taking Risk in Machine Learning

I think it comes down to thinking differently about how the government acquires things that, and one of the things that's really, really natural for machine learning or AI. The model itself is never static. And so if you realize that what you're really deploying is not the model but a pipeline that supports the model, that has embedded in it this notion that you're constantly incorporating new data to improve it. I think we should take, in some sense, those words, machine learning at heart and say this is a learning system and incorporate that in how we acquire technology.