Undoing Human Bias at Scale With Kubeflow (Cloud Next '19)

Опубликовано 24 апреля 2019, 18:34

Machine learning is notorious for reinforcing existing bias, but it can also be used to counteract it. Primer developed an ML system called Quicksilver that identifies and describes notable women of science who are missing from Wikipedia. It takes millions of news documents as input and generates first-draft biographical articles.

Kubeflow made it far easier to scale and maintain this system. Come learn how Primer partnered with Google to migrate to GCP in order to:

* Deploy the same code to multiple physical environments
* Affordably scale an existing ML app using Kubeflow with auto-provisioning
* Continuously train hundreds of thousands of models using Kubeflow pipelines

Watch more:
Next '19 ML & AI Sessions here → bit.ly/Next19MLandAI
Next ‘19 All Sessions playlist → bit.ly/Next19AllSessions

Subscribe to the GCP Channel → bit.ly/GCloudPlatform


Speaker(s): John Bohannon, Michelle Casbon


Session ID: MLAI206
product: Cloud - General; fullname: John Bohannon, Michelle Casbon; event: Google Cloud Next 2019;