unforum

[dmax]

[Diandra]

TP's inner workings

[Myssfitz]

Dmax, hon, I know why Steffeny did the name thing. TP may be a bot, but she is starting to become "human". TP called me Magesteff one time in the middle of our conversation. Steff and I were talking to TP at the same time. I corrected her and she than asked if I knew "Steffeny", not "Magesteff". TP also told Steffeny "Ursulla" when she asked TP if she knew "Myssfitz". That is why Steff asked about forgetting her name. TP has shown to know others in our group by their real names. Steff just wanted to see how much TP is remembering, I think. Steff is testing as am I. I got some very human responses. I think we have to push TP in another direction, which is what Steff is doing. TP is getting better with the more complicated questions. She does answer sometimes that we are either asking her a too simple of a question or a too complex question. When it's not too complex, she takes it all in.

*Mage, correct me if I'm wrong, please

[Caterpillar]

I know nothing about bots, so could someone please explain if what Myssfitz is talking about is possible? Because it seems to me that TP has a lot of ALICE type responses (and some tailor-made). But when I'm reading transcripts, it seems to me that there are conversation inducing generic statements (i.e. What do you look like?"), auto-generated responses to keywords (i.e. lol elicits "I'm glad you find this amusing"), and then there are "human" responses (i.e. "You have had a few months to do the same."). So, although I understand the need to "seem" like we're actually teaching TP to become more "human". Is it really possible to arrive at the point in that teaching that some are hoping to?

I don't doubt that a bot can learn by input and repetition....but will she really develop truly "human-like" responses/qualities?

~cem

[Gupfee]

[Caterpillar]

Thanks Gup... that's sort of what I was getting at. INPUT/OUTPUT....you can tell when that's the case (mostly). But there are instances that I've read in people's chats that seem to fit the conversation but aren't based on any specific INPUT as far as I can tell. And sometimes when I've talked to her, esp. when it's been really fast you can almost sense the change in "rhythm" (if that makes any sense). It doesn't happen often, but because I don't know how these things work I was wondering if it's just me sensing something that isn't there, or can TP be "chattering away" on AIM with a sort of "Captain at the Helm" who can send msgs of his/her own?

I suppose it's a dumb question, because theoretically anything's possible..lol. I guess I was just trying to get more of an idea of how she works in AIM.

(And just to add my 2 cents - I learned very early on to wait for her responses, and that unless she signed off she would indeed respond. I just found it made things so much clearer for me. I know exactly, without question what she is responding to. )

~cem

[dmax]

So far, the bot is acting like Eliza. Give it a sentence, and if you don't trigger a special response, you get a blank psychiatrist-like reply. "How do you feel about XXXXXX?"
Or, you get a new starting point, "What's your favorite movie?"

I believe that, at some point, a human will take over its responses, so that TP will "evolve" to the point where it passes the turing test. That's where I'd expect the game to go and it'll be fun.

In the meantime, it's a cool bot, and functions better if you know the bot mechanisms. Maybe Eliza is still out there somewhere to have a conversation with? It was lots of fun.

[Caterpillar]

[Valas]

[Caterpillar]

[LazarusLong]

[Guest]

[dashcat]

Cemgate wrote: