Vibeleaderboard
← All Intel
Intel / video

Don’t be data poor — Anuj Iravane, Anterior

Source
youtube.com
Author
AI Engineer
Date

Context

Anterior generates synthetic medical records by starting with the decision it wants an case to test. In Anuj Iravane's account, the company's contracts prevent it from retaining the production records its evaluation work needs. Its alternative reverses the usual decision process: choose an outcome, sample a path through an explicitly represented policy, then construct a record that should lead to that outcome.

Generation proceeds from patient-level facts to encounters and individual documents. Consistency checks look for contradictions between documents produced in parallel. Starting with an intended answer makes targeted test cases easier to construct, while those checks remain necessary to assess whether the generated record actually supports it.

Terms in this piece · Glossary
  • eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
  • inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
Read the source www.youtube.com
More from AI Engineer
Recommended reads
Comments

Checking sign-in…

Loading comments…