So we are all clear:
So, in a casual conversation about an online entity, I suggested that perhaps the entity was a "bot" that could manage a few rounds of a Truing test, and someone stated:
Obviously the statement is True, but in the process the underlying question becomes, "How many 'rounds' are required to consider the test 'successful'?" Suppose an arbitrary number of rounds is chosen....say 20 rounds....and 62 people out of 100 have no idea they had conversed with a bot rather than a human? Is this success or failure? What if the number of rounds were limited to 3 rounds and 998 people out of 1000 failed to realize they were conversing with a bot? Is this a success or failure....
So why is this important? We are all constantly bombarded by targeted advertising while we surf the web....we are at the mercy of linguistically challenged CSRs virtually every time we make a service inquiry....and if you have ever attempted to own/admin/moderate a public forum you are aware of the vast number of bots and humans a like trolling the net attempting to "make a fake post" for the sole purpose of "adding a link" to it.....In most cases a savvy surfer can simply avoid the targeted advertising....and there are tools available in most forum software packages to handle a lot of the time consuming grunt work of separating the chaff from the kernel.....but both of these situations assume bots are malevolent if not outright malicious creations....but what about the case of CSRs?
If you have ever called your credit card company, bank or a government agency, you know most have automated bots to handle the vast majority of customer needs...in this case they make no effort to conceal the fact you are "speaking with a machine", but most customers are perfectly happy to get their balance or other information in this manner....frequently diagnostic CSRs use a computer to answer questions they themselves might not know the answers to....in the case of "online chat" they frequently just "moderate the diagnostic bot"....as you describe your problem....
So let's take an example....if you enter into an online chat with a CSR and your issue is diagnosed and your problem is solved during the course of the chat and it turns out the CSR was in fact a bot.....does this pass the Truing Test? your objective was to solve your problem, not "discover anything about the source of the responses"...does this disqualify the scenario from being a Truing Test? Does the fact that you were unaware that you were not conversing with a human imply the the bot "Passed" the Truing test? Is the Truing test a measure of "intelligence", or just a measure of how useful a program can be?
I don't have any answers here, just thinking "out loud".....I have encountered some very clever bots....I think it is reasonable to assume that more intuitive and "natural acting bots" are currently under development....I don't think "all bots are bad"...some are in fact quite helpful...but what about bots that can go "20 rounds" without giving any clue that they are in fact a machine....are these bots any closer to being "intelligent" than bots like google's ad engines? Is there some "magic number or rounds" that separates intelligence from sentience? I don't know, something tells me 'no' on the sentience, but perhaps, 'intelligent' should be left open for debate.....
Please feel free to put in your $0.02 worth...there aren't any right/wrong answers AFAIK....just a thought experiment.....
Thanks!
Fish
So, in a casual conversation about an online entity, I suggested that perhaps the entity was a "bot" that could manage a few rounds of a Truing test, and someone stated:
Obviously the statement is True, but in the process the underlying question becomes, "How many 'rounds' are required to consider the test 'successful'?" Suppose an arbitrary number of rounds is chosen....say 20 rounds....and 62 people out of 100 have no idea they had conversed with a bot rather than a human? Is this success or failure? What if the number of rounds were limited to 3 rounds and 998 people out of 1000 failed to realize they were conversing with a bot? Is this a success or failure....
So why is this important? We are all constantly bombarded by targeted advertising while we surf the web....we are at the mercy of linguistically challenged CSRs virtually every time we make a service inquiry....and if you have ever attempted to own/admin/moderate a public forum you are aware of the vast number of bots and humans a like trolling the net attempting to "make a fake post" for the sole purpose of "adding a link" to it.....In most cases a savvy surfer can simply avoid the targeted advertising....and there are tools available in most forum software packages to handle a lot of the time consuming grunt work of separating the chaff from the kernel.....but both of these situations assume bots are malevolent if not outright malicious creations....but what about the case of CSRs?
If you have ever called your credit card company, bank or a government agency, you know most have automated bots to handle the vast majority of customer needs...in this case they make no effort to conceal the fact you are "speaking with a machine", but most customers are perfectly happy to get their balance or other information in this manner....frequently diagnostic CSRs use a computer to answer questions they themselves might not know the answers to....in the case of "online chat" they frequently just "moderate the diagnostic bot"....as you describe your problem....
So let's take an example....if you enter into an online chat with a CSR and your issue is diagnosed and your problem is solved during the course of the chat and it turns out the CSR was in fact a bot.....does this pass the Truing Test? your objective was to solve your problem, not "discover anything about the source of the responses"...does this disqualify the scenario from being a Truing Test? Does the fact that you were unaware that you were not conversing with a human imply the the bot "Passed" the Truing test? Is the Truing test a measure of "intelligence", or just a measure of how useful a program can be?
I don't have any answers here, just thinking "out loud".....I have encountered some very clever bots....I think it is reasonable to assume that more intuitive and "natural acting bots" are currently under development....I don't think "all bots are bad"...some are in fact quite helpful...but what about bots that can go "20 rounds" without giving any clue that they are in fact a machine....are these bots any closer to being "intelligent" than bots like google's ad engines? Is there some "magic number or rounds" that separates intelligence from sentience? I don't know, something tells me 'no' on the sentience, but perhaps, 'intelligent' should be left open for debate.....
Please feel free to put in your $0.02 worth...there aren't any right/wrong answers AFAIK....just a thought experiment.....
Thanks!
Fish
