SDS 589: Narrative A.I. — with Hilary Mason

OSEMN Process

The OSEMN data science process, developed by Hilary Mason, remains a cornerstone in the field. Initially introduced in 2010, it outlines five stages: Obtain, Scrub, Explore, Model, and Interpret. Jon Krohn highlights its widespread adoption, noting how it has become integral to data science practices today 1.

This is what you do when you are data sciencing. And it is really funny to talk about it here today, because now you look at it and you think, oh, this is so obvious. Like of course, but it wasn't obvious in 2010.

--- Hilary Mason

Hilary emphasizes the process's role in unifying diverse backgrounds within the data community, providing a common language and framework for problem-solving 1.

   

AI Development

In AI product development, Hilary Mason challenges the conventional reliance on quantitative error functions. She argues that many valuable products, like web search, operate without clear metrics of correctness 2. This perspective opens up opportunities for innovation in areas traditionally seen as too complex or undefined.

There are a large set of things we could build, but where we don't have those error metrics, and there are a ton of products that are incredibly valuable that exist today.

--- Hilary Mason

Hilary also stresses the importance of demystifying AI by focusing on its practical applications rather than the technology itself, advocating for a shift away from the hype surrounding AI 3.