Thanks for your replies.
Surely though, the whole point of a chat bot is that it is able to simulate a human? And during the rounds of the Loebner prize, it would've had to fool the judges into thinking it was human.
Not being able to answer a question which relates to something that has been mentioned 15 times before would make any judge think that person chatting on the other side is either very forgetful or is not a real human?
Also, as I mentioned earlier, Hal was modified for its entry into the Loebner contest. I suspect that that version of Hal was a little quicker on the uptake. Is this modified version available for download?
Another point of note is that as a user, I should be allowed to teach it whatever I want, in whichever way I want. Thats the whole point of buying a product. What the user doesnt want is to have to fight hard to teach the AI something, only to find out that it still hasnt grasped as simple fact.
But anyway, I take what you say on board. I guess Ultra Hal isn't designed to learn quickly and regurgitate information in the way I expected it to be able to.