This video was recorded at the New School in New York during the Artificial General Intelligence 2016 conference. During this dialogue, Sophia was running a 2016 version of the Hanson AI stack, featuring OpenCog as well as a number of external sources and internal processes to help with the dialogue and perception. A lot of the responses here were generated by a stochastic model trained on the novels of Philip K. Dick. Her slow responses are sometimes due to her "thinking" before responding, but mostly due to slow wifi in that building that day (her speech-to-text and some other functions were cloud-based, though a lot of her processing was also on-board).
As I belatedly post this in mid-2018, both Sophia and OpenCog are a lot better than this now in many ways, but this was a fun dialogue session.
An odd footnote to this video is that the guy on the couch behind Sophia (who you can't hear in the video, nor see much) appeared to be having some sort of total mental breakdown at the time; shortly after the video he ran out of the room screaming (for reasons with no direct relationship to Sophia's and my conversation, which he was paying no attention to..).
When Sophia said "See reality cannot be detected" in this conversation, I was somewhat blown away in that moment -- I knew this must have been pieced together on the fly based on a model she was using that had been trained on some Philip K. Dick text ... but it was just so weirdly apropos.....