Submitted by Alex Scott on Tue, 14/11/2006 - 09:36.
...I was thinking about memory, and how lending the program with some kind of memory could make interraction with it far more convincing and lifelike. One of the real turn-offs with chatbots is when they make it abundantly clear to you that (like people in real life) they are not really responding to anything you are saying... they are not adapting to what you are putting out there. So interractions tend to be either random or annoyingly circular. There is no sense of give and take. But for this, the program would have to be able to differentiate between users.
This need not be very complex. Currently, many users of the site must log on anyway. All the program need do is keep track (for a limited period of time, we could set the memory horizon at a month in the past or something) of what it had already said to whom (to avoid repitition of feedback) .
More Random Ideas 01
...I was thinking about memory, and how lending the program with some kind of memory could make interraction with it far more convincing and lifelike. One of the real turn-offs with chatbots is when they make it abundantly clear to you that (like people in real life) they are not really responding to anything you are saying... they are not adapting to what you are putting out there. So interractions tend to be either random or annoyingly circular. There is no sense of give and take. But for this, the program would have to be able to differentiate between users.
This need not be very complex. Currently, many users of the site must log on anyway. All the program need do is keep track (for a limited period of time, we could set the memory horizon at a month in the past or something) of what it had already said to whom (to avoid repitition of feedback) .