[DemonRaven]

Is it just me or does some of the people at ai_zone forum seem to be a tad testy when you bring up needing to do things different as far as chatbots go?  I think that there needs to be a different way of doing it as tossing facts at chatbots does not necessarily improve their conversational skills. There needs to be a way to give them common sense. Without it all the facts in the world won't make them better.

The very definition of  a chatbot is a program that talks like a person. IT carries on a conversation. It is not necessary for it to know everything. IT is necessary for it to somehow understand what we mean and the nuances of speech. Joy hardwood who created God Louise and i a long time ago spoke about this very thing. We agreed that something different needed to be done. There needed to be a new way to instill common sense to them. How that would be accomplished i am not sure but i think it some how needs to be close to it having true intelligence of some sort. Even animals can communicate with each other. I think that when we finally understand how the brain works that then we might be able to duplicate that in a computer program. Until then we are kinda swimming in the dark about it.

But it does not mean we should not keep trying because if nothing else chatbots are fun to play with.

[Art]

I've sort of thought about the same thing. Chatbots should Chat! Most people I chat with don't know everything there is to know and even if they came close it wouldn't improve the chat relationship (maybe no one wants a "know-it-all").

There are far too many programs and now hardware devices to do all those "other things" in our life that require turning things on / off or bright / dim or whatever....

How about a chatbot that can respond to emotions within the context of speech (spoken or typed)? How about a chatbot that can have and develop a personality over time and become engaging with the user? How about a chatbot that can be a friend and remind the user of a few things but again, not that knows the rotation of the Earth at the equator!? That's not important!

How about a chatbot that chats with a level of machine/software intelligence that has a grasp of English and does not confuse I/me/you posession?

How about a chatbot that learns and remembers everything (within reason) taught to it through conversation or specialized learning processes?

Perhaps one day soon....

[Don Patrick]

How that would be accomplished i am not sure but i think it some how needs to be close to it having true intelligence of some sort.

People are never grateful if you point out their flaws without adding constructive solutions. Chances are most chatbot creators are aware of the weaknesses of their approach, but just aren't Phd computer scientists that would know how to solve the greatest challenge of several millennia.

If you want my opinion, a conversational approach should have an expert on social interaction, psychology and programming on board, all in one person, because psychologists and programmers go together like cats and dogs. For starters there is such a thing as semantic analysis, which can be used to gauge the user's emotions and preferences. Your best bet there would be Bruce Wilcox' creations. Secondly you'd need a well developed system of social rules that are not present in current chatbots. To be honest Eliza is still the best at this for abiding one rule: Show interest in the user. The rule Eliza broke however is that she didn't remember what the user said before, so you're also going to need a decent knowledge database and a learning mechanism. Third is context, without which half of human conversation makes no sense. There aren't any particularly good solutions to this yet invented beyond word disambiguation. And lastly a thorough understanding of language, again, not something science has yet solved well. Some scientists are working on Winograd Schemas methods to allow a natural use of "he/she/they" without confusion and it's not the hardest trick in the book for regular conversation, but nobody is integrating it in a conversational engine because there is too much money to be made with it in other fields. More feasably, Stanford's grammar parser is freely available to all, but for some reason very few chatbot creators use it. Perhaps they don't know how to use it. Why don't you?

As for myself, I just pursue other goals. Any improvement to my program's conversational engine is a pastime only. I'm more interested in getting the intelligence and the necessary language skills down before social interaction.

[Don Patrick]

If you are referring to Fariz' use of exclamation marks in response to you though: That is enthusiasm, not a negative attitude. He was just correcting your misunderstanding of Tensorflow. The problem with Google's kind of machine learning I can sum up in one word: Cleverbot.

[DemonRaven]

To be honest I was not aware of The Stanford Parser. Now that you have pointed it out I will check it out.

As far as my comments I am not faulting anyone person. I just think that we need to try something else.

[DemonRaven]

To be honest I am elderly so a lot of stuff I have forgotten how to do. I am also finding myself running into road blocks in my mind and am finding myself getting distracted easily and my mind wandering. So it is harder to learn new things. But I am keeping at it anyway and will figure out the The Stanford Parser eventually to use it even if it means kicking furniture in frustration at time lol.

[Don Patrick]

Pretty much everyone except (or including) chatbot creators themselves thinks it takes a better approach than pattern matching.
The thing about using Stanford's Parser is that you're going to have to learn to program. I assume this is why the average home-based chatbot enthusiast doesn't even try. Alternatively you could take a look at the grammar parsing functions in ChatScript.

Grammar parsing won't directly cause better conversation either, but what it does is add a degree of accuracy to the computer's understanding. Rather than pattern matching that responds to the keyword "yard", the parsed grammar of a sentence can tell you that the noun phrase "in the yard" is just a location specifier and not a good prompt to start yapping about yards, when the rest of the sentence says "I buried my cat in the yard". Instead, the subject, verb and object are the important parts to respond to.
The difference isn't that pattern matching can't do the same, but that a grammar parser does it all automatically and consistently for every sentence. It also handles ambiguity and can resolve pronouns, no matter which words the user typed.

The interesting part is of course what you do with it once you've pinpointed the important elements of the user's input. I mentioned semantic analysis before. Using this you can have the computer check whether "bury" is a negative event, and then it knows that it should respond with sympathy, or apology, or take insult, depending on what was done to who. Again, all this without having to manually single out specific words to respond to.

This reminds me of Tron's "Bit", who could only say "yes" or "no", but it was meaningful because it understood the question.

[DemonRaven]

I am not naive about programming Don. I created my own website. Unlike some people I dislike the ready made ones as there is not enough variety in them to truly personalize it. I do not even like using software that makes it "easy" to code a website for the same reason. There is not enough wiggle room to personalize it. I am one of those weird persons who likes to code by hand and only using basic html editors that just help out as far as not having to do as much writing.  I have been around the chatbot community and worked with and/or looked at the code of almost every chat bot there is out there. I even know enough that i can usually find a way to look at programming that creators try to hide.   I am not new to any of this. I have been around since 2000  and my website has been in existence in one form or another since 2002. My websites name has been in existance longer then your precious chatbot.org which I do remember when it first started up. I in fact at the time was irritated that because they had more money at the time to afford a domain name and i could not  and back then it gave them a advantage over my website. The only reason I seem to be new is that i don't use my real name as a user name any more.

I also unlike some people do not claim to know it all or think I am some how more intelligent then they are. I never claimed to have all the answers. I also never claimed to have "the" answer to the riddle of the puzzle we call Artificial Intelligence aka cognitive programming. It however does not take a rocket scientist to figure the basic flaws of chatbots which are painfully obvious.

The only thing about me that is new is that I recently came back to the AI community after several years of absensence and forgot the dang password at the other site.  A user name which i have had there for several years now. I contacted at least 4 of their board members and got no response to update it so i gave up and found a way to work around it and create a "new" user name.

So yes I consider myself a amature because I am not a IT programmer. I do however understand the basic concepts of coding. I also have been teaching myself more advanced website coding because it is the key to eventually learning the harder programming languages.

[DemonRaven]

Computers in general lack common sense. There are some humans that are almost as bad my father being one of them. The higher the IQ the worse the social skills seem to be.

[DemonRaven]

Frankly considering the fact that the way my mind works now, resembles adhd more then any thing else i think i do a damn good job of learning new things.

[DemonRaven]

This is who I am. I am actually searchable:

[pandorabots-general] Re:contest - Archivum.info
www.archivum.info/...com/.../(pandorabots-general)-Re-contest.html
Apr 21, 2005 - from [sheryl clyde] ... with whatever contest or contests you choose to enter. sheryl clyde Chat Bot Friends http://chatbotfriends.com ----- Original ...

From info@pandorabots.com Fri Jan 17 06:01:52 2003 ...
lists.pandorabots.com/pandorabots-announce/2003-January.txt
Jan 17, 2003 - ... 17 06:01:52 2003 From: info@pandorabots.com (sheryl clyde) Date: Fri ... my website: http://www.chatbotfriends.com that they can > send me ...

Frankly i am getting fed up with the patient little verbal indulgent pats on the head because people think i am new.

[Freddy]

Sorry if I've offended you in any way Sheryl. To be fair though I don't think our paths had crossed until you joined here. I don't go to many other AI forums these days. The web is a big place I don't know everyone yet 

If I've said anything about your work, my comments are genuine, I'm not being funny about it

[Don Patrick]

DemonRaven; Kindly blow off your steam in another direction. Your elaboration only confirms my assumption that you are new at A.I. programming, plus you were using your age as an excuse for not being able to learn well. So I feel no need to apologise for my cautioning. But I'll gladly stop trying to be helpful if you prefer that.

On a minor note, I am not particularly affiliated with chatbots.org

[DemonRaven]

IS there a way to block people on here.

[Data]

IS there a way to block people on here.

In a word, "no".

It would be better if we could all just get along and respect each others comments and points of views.
We don't have to agree with them but everyone has a voice here and as long as no one gets abusive or nasty everything is fine.

I've been through this thread and as far as I can tell no one has crossed the line ... yet.