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

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.

Patrick Kulp
@patrickkulp patrick.kulp@adweek.com Patrick Kulp is an emerging tech reporter at Adweek.
Recommended articles
You Might Like

How the Latest Identity Solutions Stack Up to Consumer Privacy Demands

By Ian Meyers

3 New Behaviors of Media Consumers and What to Do About Them

By Zephr

While Talk of the Post-Pandemic Future Is on the Rise, SMBs Continue to Struggle

By Lorenzo Papa, Chief Revenue Officer, Captivate

Why Print Matters More Than Ever in a Post-Pandemic World

By Tonya Powers, Director of Marketing Production Print Solutions, Canon Solutions America

{"channel":"elasticpress","title":"Recommended","post_types":["post","page"],"lineup_id":"54851","since_qty":30,"since_time":"d","until_qty":0,"until_time":"m","q":"","num_posts":0,"vertical":"performance-marketing"}