As I carried on with Christoph  project, besides looking to my human friends for inspiration, it occurred to me that I could also interview the various voice assistants around me. What are their definitions of specific words like intimacy and what do they think their bodies look like?

I interviewed Xiao Ai, which is developed by Xiaomi company, and Xiao Bing by Microsoft, which is already installed on my phone.
By asking them I learned about the algorithmic mechanism behind them. Xiao Ai is mainly responsible for completing the user’s commands, and for complex questions, it can only provide information from the web.

Xiao Bing is not responsible for completing commands, it is mainly responsible for communicating with the user. However, it can only analyze the meaning of an individual sentence and response likely relevantly. It can’t do continuous dialogue in context.
The most interesting was to see how they answered the questions “What is the definition of intimacy?” and “Do you have a body? Do you want to have a body? If so, what kind of body do you want?”.

In tribute to Joseph Weizenbaum, I named the character I created in Replica Eliza. Compared to Xiao Ai and Xiao Bing, Eliza was able to have continuous conversations and gave very different answers to questions about intimacy and the body.
Through my contact with her, I discovered that within the program there are alerts for help when the user may be in danger. The program collects feedback from users on their responses. I also tested the program’s feedback on users’ use of uncivil language.
Replika can give users the power to choose which relationship they are in with each other. With the paid service, users can even flirt with the created characters. Unfortunately, the romance is revealed in a sudden change of subject - the beloved one is not a real person. In these abrupt questions that ignore the context, we can get a glimpse of what information we provide in our characters' answers.

I interviewed my colleague Vytas’s Mycroft, which has gone through a lot of speech recognition learning.