Q No, I think it's remarkable. For disclosure, I worked here for four years, um, and was really quite impressed. Um, anyway, how does the, the handoff work between the, the machine and the humans? You mentioned the human in the loop. How does that work practically?
A Yeah, so when you have every one of our machine learning based applications has a different trade-off between precision, recall, and speed. Um, and it's really not optimize F all the time or optimize latency all the time. It's a different, it's a different trade-off for each one. When we have time or offline if, if necessary due to latency constraints, we use an assortment of active learning methods. Active learning? Active learning methods. Um, and we have a team of data scientists Some of them with very specialized expertise in finance, and some of them with more generic expertise who help us to label data, who help us to design the algorithms, who provide subject matter expertise, um, so that we can continually improve all of our systems. And that's something that historically has been done mostly with rule-based systems, but today we do it with, um, deep learning and a lot of decision trees.
AI assessment note: “help us to label data, who help us to design the algorithms”