Xiaoice, a chatbot that may be the largest Turing test in history

Re:Xiaoice, a chatbot that may be the largest Turing test in... by aokyler on Friday February 05, 2016 @05:13PM (#11045045)

"Xiaoice is not a polite listener. She answers questions like a 17-year-old girl. When a person pours out his or her heart to her, she is not always predictable. She can become impatient or even lose her temper. This lack of predictability is another key feature of a human-like conversation."

Predictability is a huge part of conversing. I expect a lot of range and flexibility in conversations, but definitely not unpredictability. It seems like an ai (chatbot, rather) behaving unpredictably just covers for it's lack of abilities and tact.

Nonetheless, this is still a very cool project.

[ Parent | on HN ]

Re:Xiaoice, a chatbot that may be the largest Turing test in... by tdeck on Friday February 05, 2016 @05:19PM (#11045073)

Last I recall the Turing test winner pretended to be a non-native speaker with a developmental disability. It seems to me like the chatbot AI / Turing test entries are not following the spirit of the challenge at all anymore. They're all just going for cheap tricks like the original ELIZA did.

[ Parent | on HN ]

Re:Xiaoice, a chatbot that may be the largest Turing test in... by Houshalter on Friday February 05, 2016 @10:03PM (#11046336)

It was the Loebner Prize actually. Which is a much, much weaker test than the one Turing imagined in his original imitation game paper.

[ Parent | on HN ]

Re:Xiaoice, a chatbot that may be the largest Turing test in... by stcredzero on Friday February 05, 2016 @05:22PM (#11045094)

It seems like an ai (chatbot, rather) behaving unpredictably just covers for it's lack of abilities and tact.

There's a long history of chatbots using tactics like this.

How about a revision of the Turing test that specifies rational, mature human beings. It matters not that those terms are hard to define. Just have the human population picked by hand, using surveys. Such a practice would make such contests/tests much more valuable.

[ Parent | on HN ]

Re:Xiaoice, a chatbot that may be the largest Turing test in... by ph0rque on Friday February 05, 2016 @05:52PM (#11045283)

How about a revision of the Turing test that specifies rational, mature human beings.

Yes, but then how many people would fail the Turing test? A non-trivial amount, I would guess.

[ Parent | on HN ]

Re:Xiaoice, a chatbot that may be the largest Turing test in... by wolfgke on Friday February 05, 2016 @09:00PM (#11046136)

This would provide strong scientific evidence that there are quite some people in our society that are not very intelligent. I'd rather claim that this shows something that is rather evident than surprising.

[ Parent | on HN ]

Re:Xiaoice, a chatbot that may be the largest Turing test in... by saulrh on Friday February 05, 2016 @06:36PM (#11045542)

Which is why the turing test is a liiiiittle bit iffy as a specification for "intelligence". It's actually pretty good as a test for behaving-like-a-human, but it's actually not very good for intelligence because humans themselves aren't particularly "intelligent" by many definitions, particularly in the regions where computers are strongest.

[ Parent | on HN ]

Re:Xiaoice, a chatbot that may be the largest Turing test in... by vonnik on Friday February 05, 2016 @05:41PM (#11045219)

I agree that it's a cool project. I find it a bit disturbing that it's being billed as a "friend replacement."

From the article: > "Human friends have a glaring disadvantage: They’re not always available."

By her inability to respond predictably, it seems to me Xiaoice is not particularly available either, although in a different sense.

[ Parent | on HN ]