DBTB INT Vita Markman
summary
At LinkedIn the central claim is that machine learning is not magic and supervised models require clean and consistent data so expect practical rules not theory. The talk stresses hands on vetting: label audits, simple heuristics to catch noise, and manual fixes when crowdsourced labels fail. It argues that accuracy gains come from better input not bigger models and documents repeated failures where unvetted data caused models to collapse in production.
AI-written summary · about
chapters (4)
related talks
31mReliable Data for Large ML Models: Principles and Practices
21mDataEngBytes 2023 - SYD-T2-04 - Suneeta Mall
38mAutomatic DB annotation with ML tools and its problems
04mManaging Machines
04mInterview
49mGetting to True Predictive Risk: Will Data Accuracy Thwart AI’s Potential?
27mDataEngBytes 2023 - BRI-T1-07 - Joe Reis
07mPredictive Modeling and Machine Learning