Talking to myself: how I trained GPT2-1.5b for rubber ducking using my chat data

Re:Talking to myself: how I trained GPT2-1.5b for rubber duc... by qnxub on Thursday January 23, 2020 @05:30PM (#22132990)

Is this the best way to create a chatbot with my personality? I feel like I would want to fine tune some things so it is giving real responses about my preferences, hobbies, etc.

My use case is preserving my personality for loved ones after I die.

[ Parent | on HN ]

Re:Talking to myself: how I trained GPT2-1.5b for rubber duc... by backupcavalry on Thursday January 23, 2020 @10:44PM (#22135221)

Not knocking you but I'd love to see some research on whether this would actually be a positive for loved ones - for me, I know I'd prefer them to move on to fresher things in life.

That and the Black Mirror episode another commentor mentioned.

[ Parent | on HN ]

Re:Talking to myself: how I trained GPT2-1.5b for rubber duc... by Coviam on Friday January 24, 2020 @03:37AM (#22136448)

With multilingual chatbots, you can reach customers across the globe, enhancing your customer support and the user experience. With the best AI chatbot, it is possible to build a chatbot that can do so. Learn more: http://s.engati.com/1ql

[ Parent | on HN ]