Just wondering, what do you guys think it takes to create a system that in a blind test will convince a person communicating to it that it is in fact another human being.
Winning the Loebner Contest
- Thread starter Promethply
- Start date
Yes, exactly like what the article described.
It seems that creating a system that can respond and interact normally with a human being is a great challenge, because the machine can only use a fixed number of programmed responses.
So how do we create a "chatbot" that can communicate to us like as if we are using Windows Messenger and chatting to a friend, or at least another human being.
It seems that creating a system that can respond and interact normally with a human being is a great challenge, because the machine can only use a fixed number of programmed responses.
So how do we create a "chatbot" that can communicate to us like as if we are using Windows Messenger and chatting to a friend, or at least another human being.
Hmm..i am not a programmer, nor one that knows enough in this type of area to comment....but anyway lol
If you could program a computer to do this:
Record an infinite number of Internet conversations each day(if ppl wanted to donate to the project ect, and let a comp tape their convo's)
and then have the computer sort them and be able to use them as repsonces, for example when the computer see's "this" then it responds with the computer recorded and determined as the most common responce for that particular "this"....sorry i am having trouble articulating this for some reason....after a month or even a week the computer would have a very large amount of preprogrammed responces...of course its not that simple(spelling mistakes...ect...ect)...but i think it could be done if you get what i am saying
If you could program a computer to do this:
Record an infinite number of Internet conversations each day(if ppl wanted to donate to the project ect, and let a comp tape their convo's)
and then have the computer sort them and be able to use them as repsonces, for example when the computer see's "this" then it responds with the computer recorded and determined as the most common responce for that particular "this"....sorry i am having trouble articulating this for some reason....after a month or even a week the computer would have a very large amount of preprogrammed responces...of course its not that simple(spelling mistakes...ect...ect)...but i think it could be done if you get what i am saying
Something like that would be like "brute force", still probably not too great.
You really need to code an expert knowledge of language and logical thought processes to use it. I'd say the coding skill is the main thing standing in the way, maybe power too. This would go hand-in-hand with creating a better AI.
Really appreciate all of your input, and must apologise that I couldn't get back sooner, as I've got a job deadline looming.
lol np...we all have work...this is our little retreat lol
Yeah...but offer a free NSync t-shirt for all those teenyboppers out there who donate and have the most IM's.....lol you could get a whole slew of em.....but yes it would take a lot to code to get the computer to understand and respond....it would be hard just for the abbreviations alone lol
Yeah...but offer a free NSync t-shirt for all those teenyboppers out there who donate and have the most IM's.....lol you could get a whole slew of em.....but yes it would take a lot to code to get the computer to understand and respond....it would be hard just for the abbreviations alone lol
Conversation without context wouldn't be very convincing. For example -
What did you think of the big game on the weekend ?
DrPizza
Administrator Elite Member Goat Whisperer
Semi-serious response:
if the computer responds with phrases like
"wut R U doing?"
"wut color is ur house?"
Then, the person interacting with the program might be more easily convinced it's human - just a stupid kid.
Seriously, though, part of the difficulty will be in designing how "intelligent" your simulated human speech is. I think it'd be far more easy to pretend to be a 12 year old and carry on a conversation with a 12 year old - rather than trying to interact with, for example, a newspaper reporter (too much in the current events category), an astrophysicist, or any other person with a more sophisticated vocabulary and knowledge.
if the computer responds with phrases like
"wut R U doing?"
"wut color is ur house?"
Then, the person interacting with the program might be more easily convinced it's human - just a stupid kid.
Seriously, though, part of the difficulty will be in designing how "intelligent" your simulated human speech is. I think it'd be far more easy to pretend to be a 12 year old and carry on a conversation with a 12 year old - rather than trying to interact with, for example, a newspaper reporter (too much in the current events category), an astrophysicist, or any other person with a more sophisticated vocabulary and knowledge.
Gibsons
Lifer
- Aug 14, 2001
- 12,530
- 35
- 91
hmm, so suppose your part of the experiment as a test subject. You're put in a chat room with an unknown entity.
What sort of queries would you use to try to figure out if your chat partner was a bot or human? Or, what would you do to convince your chat partner you were human? You might be able to convince them you're a bot by typing things like error:could not parse line 26:"stupid_answer_to_stupid_question_0122389#######."
What sort of queries would you use to try to figure out if your chat partner was a bot or human? Or, what would you do to convince your chat partner you were human? You might be able to convince them you're a bot by typing things like error:could not parse line 26:"stupid_answer_to_stupid_question_0122389#######."
Actually, the chatbot will get away with that kind of conversation, as long as
it's the chatbot who's asking questions similar to the above question.
But if it is the real person asking the chatbot that question, than up to now,
the chatbot would fail.
It's very difficult to simulate an average conversation, especially ones that has no purpose.
also not every single person knows everything either....it would be possible to have smart preprogrammed responces to questions the computer does not know...ex
Person: what did you think of the game last night in NY...
*the computter hasnt received any date on this, but does recognize game last night, which is in reference to scores...."
Computer: i did not see it i was tired, whatdid happen??
And doing what i suggested could be a start...while using that data the computer could build itself an artifical AI with what it sees and the past experience of convo's in its past....??
Lol about context....why not have it tape these forums..lol ;-))
Person: what did you think of the game last night in NY...
*the computter hasnt received any date on this, but does recognize game last night, which is in reference to scores...."
Computer: i did not see it i was tired, whatdid happen??
And doing what i suggested could be a start...while using that data the computer could build itself an artifical AI with what it sees and the past experience of convo's in its past....??
Lol about context....why not have it tape these forums..lol ;-))
I don't think it could actually. Real people have difficulty faking a coherent conversation about sports they know nothing about, a chat bot which doesn't understand the concept of sports, how the statement relates to that concept or how the terminology is strung together doesn't have a chance.
Not to mention sarcasm, irony, figurative speaking and allegories are essentially the kiss of death 8)
What kind of response are you going to come up to a statement like "Tiger woods has more chance of winning the superbowl this year than the redskins do" ?
thing is though...if your using statements that can trip up plenty of REAL human beings that do not know....how is that fair for a test....lol
I bet we can all think fo that special someone at the office who would try to respond to that
I bet we can all think fo that special someone at the office who would try to respond to that
DrPizza
Administrator Elite Member Goat Whisperer
Exactly
I wonder how many humans would fail a Turing test. "subject lacks human intelligence and capability to carry on a conversation."
AFAIK the conversation carried out during the test would be similar to ones carried out between strangers,
and this is probably why it is so difficult to program chatbots to respond in a humanlike manner --
since there's just too many possible questions and appropriate responds to these questions.
Any suggestion as to which Computer Language we can use to create the system?
Thanks again, for your suggestions and ideas guys.
and this is probably why it is so difficult to program chatbots to respond in a humanlike manner --
since there's just too many possible questions and appropriate responds to these questions.
Any suggestion as to which Computer Language we can use to create the system?
Thanks again, for your suggestions and ideas guys.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 23K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 19K
-
-
-
Facebook
Twitter