4 Steps To Bringing AI-Powered Products To Market: How Lyft Builds Intelligence Into Product

4 Steps To Bringing AI-Powered Products To Market: How Lyft Builds Intelligence Into Product 960 640 C-Suite Network

Lyft’s head of machine learning Gil Arditi talked today at VB Summit 2017 about how to take AI out of the lab and into real-world products that provide competitive marketplace advantage.

Lyft, which just closed an additional billion-dollar round of financing led by Google, has an internal four-stage process, Arditi says.

(Photo by Kelly Sullivan/Getty Images for Lyft)

The first step, of course, is data. And it’s not just about the amount.

“You have to normalize the data, make the data very clean, and very standard,” Arditi said.

This is a critical first step not only because machine learning requires data to process, but that bad data can lead to sub-optimal decisions that actually kill competitive advantage. Neural networks in particular are very sensitive to noise in the data, Pinterest’s head of engineering Li Fan has…