I just had a horrifying realization about Elon's grand plans here.
They're going to do a join on a several different data sets with different collection latencies, data quality practices, etc. They're going to act like all the data is perfect, the schema all match, there are no special cases...
They're going to do a join on a several different data sets with different collection latencies, data quality practices, etc. They're going to act like all the data is perfect, the schema all match, there are no special cases...
Comments
The world's worst case of "did you even read the documentation"?
Whatever AI they feed it into will also just assume data is perfect
Buddy, that’s a stable system that underlies the largest single health-insurance payments processing in the world. It’s 4% of the GDP. You don’t “just” anything with it. You have to fix your own assumptions instead.
• assume datasets' granularities match
• assume id columns can be treated like primary keys
• assume different datasets are equally validated/reliable
• all sorts of false assumptions about datasets' cardinality
• etc
• etc
• etc
ICYMI (https://www.wired.com/story/null-license-plate-landed-one-hacker-ticket-hell/)