Artificial Intelligence

Microsoft Warns Its Advanced Chatbot Might Say Offensive Things

Software giant releases blueprint for DialoGPT, its Reddit-trained AI

Purple blueprint robot with neon grid chat boxes coming out of its mouth
The GPT-2 was trained on hundreds of millions of Reddit posts created between 2015 and 2017.Getty Images
Headshot of Patrick Kulp
By Patrick Kulp

November 8, 2019

Microsoft researchers have released a chatbot version of a cutting-edge text generator trained on tens of millions of Reddit posts—albeit with a disclaimer in place should things get offensive.

AW+

WORK SMARTER - LEARN, GROW AND BE INSPIRED.

Subscribe today!

To Read the Full Story Become an Adweek+ Subscriber

View Subscription Options

Already a member? Sign in

Headshot of Patrick Kulp

Patrick Kulp

Patrick Kulp is Adweek's senior reporter covering emerging tech.

Recommended articles