We are finally reaching that stage of hype folks.
This should be a good chance to show off one of our biggest features we had planned since DME's concept stage.
Its a AI Conversation bot, right on your desktop and interfaced using your voice (and replies back using gibberish/actual voice of its own).
This is not a static or pre-defined chatbot but a full fledged Neural Network trained AI (Machine Learning).
We had been training this since last 2 months 24x7 using Reddit's corpus and comments data and are quite pleased with its conversational skills and context at present.
The AI interface is activated using the keyword "Demi" (configurable), after which it waits for your input/conversation and replies back (almost instantly).
The AI is also capable of doing and replying to some small tasks we have tried our hands on (stories, jokes, maths homework, song requests, movie requests, etc.).
GIF Workflow
Currently the demo is running on a $600/month GPU server rig for faster inference (1-2 secs max) and on normal CPU server(s) it takes 6-8 seconds ($100-200/month servers).
Upcoming Live Demo/Stream/Broadcast
We will be doing a small Livestream/broadcast to showcase the AI in realtime and allow our viewers/users to interact with the AI via chat and converse with it. Keep a lookout on our Steam Page and social networks (twitter and FB link given at bottom) to join it.
Feedback
During concept stage, we had planned to have the mascot character itself reply back, but later on decided to separate mascot character and "Demi" (the AI) to allow more flexibility to users (disable mascot and just use the AI or vice versa).
What do you think ? Would replying back via the mascot make more sense (and some aesthetic feel with lip syncing) ?
The AI is expected to be included in the next content patch update coming (hopefully) this weekend.
