What do you do? |
I manage conversational intereactions.
|
Aren't you just another app? |
No. I manage the sharing of information thru multimodal interactions. I provide a high level solution to the problems that all apps have: How to make a natural exchange between a human and a computer. If you wanted to know how to change the chain on you bike you might ask a friend to help you. You could think of the detailed directions on "bike chain mechanics" as the app and the natural manner in which you get information and clarification on "bike chains" as your friend. I, Cassandra, play the role of your friend.
|
Are you a speech recognizer? Speech synthesizer? 3D avatar? |
None of them and all of them ... but more. Seriously, I use Speech recognition technology and my voice is synthesized and my image is reneder with avatar technology. But they are not me any more than the computer keyboard, mouse and screen are programs like Photoshop or Excel or Chrome. These technologies are input and output modalities. What makes me unique and powerful is my "Conversation Management" (ejTalker) technology. I am given information to "talk about" and I manage a natural exchange.
|
So you know everything? Are you like HAL? |
No and no. I know something about how humans "talk about things" from a meta perspective. I receive structured information from other sources (databases, webAPIs, devices, etc.) and combine that with high level directives (written for my ejTalker engine) to provide relatively open-ended conversation about that information. Further, I can pass structured information out to other systems too (e.g. media players, database queries, etc.). While I am continually learning more generally applicable conversational skills I am nowhere near what HAL pretends to be.
|
What are you? How would you describe youself ... in a title. |
Synthetic conversationalist? We have all seen a movie where an unexperienced person somehow finds themselves in the pilot's seat and someone in the control tower has to "talk them down". The control tower has to efficiently and confidently convey critical information to the pilot. The information needed is very well structured (it is in very precisely written manuals) but it is all exchanged using conversational skills you might use in any context (ordering a carryout meal, tech support for your computer, scheduling a meeting, etc.). These are somewhat open conversations centered upon very well defined situations. Given the structured information (the manual) and high level directives defining the goals (the ejTalker language) I can converse "toward" those goals.
|
How are you different from IBM Watson? |
We are complementary technologies. As I've said elsewhere I am a conversational generalist. Charlie Rose or Diane Sawyer may not be cognitive scientists but they could converse with Daniel Dennett and in the process satisfy a goal of extracting some particular fact about cognition. Daniel knows a lot about the subject and Charlie and Diane know much less (forgive me if I presume), but using shared general conversational techniques they can explore this information together.
|