 | I believe that might be still a little beyond what Hal can do for now.
As far as I understand, the current hal version can only learn from a 'right' and 'wrong' pattern. "Right" being when you continue the conversation, reinforcing what Hal said. And the usual "wrong" command.
It could, without doubt, be trained to ask what the unknown (x) was. But it would not be able to learn the definition of (x) unless that definition was taught through the 'wrong' command.
|  |