rw-book-cover

Metadata

Highlights

the approximate qualitative averages that get built by these models bulldoze bad data in ways careful mathematical computations do not. (View Highlight)

We haven’t gotten twitchy about broken dashboards because, frankly, people don’t get twitchy about broken things that they don’t care about.7 (View Highlight)

If data quality matters more now than it did before, it seems like that will be the reason why: Not because AI models care more about higher quality data than traditional reporting, but because, at least during the current AI circus, businesses care more about AI models than they ever did about reporting. (View Highlight)

Consider, for example, the oddity of the entire data contract debate. If data teams need people to sign actual contracts so that they’ll be motivated to give us good data—good data that we are supposed to manufacture into something useful and give back to them—how valuable can that thing that we’re making really be? You only need hall monitors when nobody believes that the rules actually matter. (View Highlight)