First, is there an agreed upon definition for person? We need to define that and make sure we agree before we should go much further, but I’ll give it a try anyways.
All Turing tests are not intuition pumps. There should be other Turing tests to recognize a greater degree of personhood. Perhaps if the investigator can trigger an existential crisis in the chatbot? Or if the chatbot can be judged to be more self-aware than an average 18 year old?
What if the chatbot gets 1000 karma on Less Wrong?
It seems like this idea has probably been discussed before and that there is something I am missing, please link me if possible. http://yudkowsky.net/other/fiction/npc is all that comes to mind.
I think I’m confused: what I assumed you meant was a chatbot in the sense of ELIZA (a program which uses canned replies chosen and modified as per a cursory scan of the input text). Such a program is by definition not a person, and success in Turing tests does not grant it personhood.
As for my second sentence: Turing’s imitation game was proposed as a way to get past the common intuition that only a human being could be a person by countering it with the intuition that someone you can talk to, you can hold an ordinary conversation with, is a person. It’s an archetypal intuition pump, a very sensible and well-reasoned intuition pump, a perfectly valid intuition pump—but not a rigorous mathematical test. ELIZA, which is barely clever, has passed the Turing test several times. We know that ELIZA is no person.
Sorry, by chatbot I meant an intelligent AI programmed only to do chat. An AI trapped in the proverbial box.
I agree that a rigorous mathematical definition of personhood is important, but I doubt that I will be able to make a meaningful contribution in that area anytime in the next few years. For now, I think we should be able to think of some philosophical or empirical test of chatbot personhood.
I still feel confused about this and I think that’s because we still don’t have a good definition of what a person actually is; but we shouldn’t need a rigorous mathematical mathematical test in order to gain a better understanding of what defines a person.
The Turing test isn’t a horrible test of personhood, from that attitude, but without better understanding of ‘personhood’ I don’t think it’s appropriate to spend time trying to come up with a better one.
No. The Turing test is an intuition pump, not a person-predicate.
First, is there an agreed upon definition for person? We need to define that and make sure we agree before we should go much further, but I’ll give it a try anyways.
All Turing tests are not intuition pumps. There should be other Turing tests to recognize a greater degree of personhood. Perhaps if the investigator can trigger an existential crisis in the chatbot? Or if the chatbot can be judged to be more self-aware than an average 18 year old?
What if the chatbot gets 1000 karma on Less Wrong?
How would you Turing test an oracle chatbot? http://lesswrong.com/lw/1lf/open_thread_january_2010/1i6u
It seems like this idea has probably been discussed before and that there is something I am missing, please link me if possible. http://yudkowsky.net/other/fiction/npc is all that comes to mind.
I think I’m confused: what I assumed you meant was a chatbot in the sense of ELIZA (a program which uses canned replies chosen and modified as per a cursory scan of the input text). Such a program is by definition not a person, and success in Turing tests does not grant it personhood.
As for my second sentence: Turing’s imitation game was proposed as a way to get past the common intuition that only a human being could be a person by countering it with the intuition that someone you can talk to, you can hold an ordinary conversation with, is a person. It’s an archetypal intuition pump, a very sensible and well-reasoned intuition pump, a perfectly valid intuition pump—but not a rigorous mathematical test. ELIZA, which is barely clever, has passed the Turing test several times. We know that ELIZA is no person.
Sorry, by chatbot I meant an intelligent AI programmed only to do chat. An AI trapped in the proverbial box.
I agree that a rigorous mathematical definition of personhood is important, but I doubt that I will be able to make a meaningful contribution in that area anytime in the next few years. For now, I think we should be able to think of some philosophical or empirical test of chatbot personhood.
I still feel confused about this and I think that’s because we still don’t have a good definition of what a person actually is; but we shouldn’t need a rigorous mathematical mathematical test in order to gain a better understanding of what defines a person.
The Turing test isn’t a horrible test of personhood, from that attitude, but without better understanding of ‘personhood’ I don’t think it’s appropriate to spend time trying to come up with a better one.