4 ms·
"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
by aokyler 11y ago
"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.
- tdeck 11y agoLast 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.
- Houshalter 11y agoIt was the Loebner Prize actually. Which is a much, much weaker test than the one Turing imagined in his original imitation game paper.
- stcredzero 11y agoIt 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.
- ph0rque 11y agoHow 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.
- saulrh 11y agoWhich 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.
- wolfgke 11y agoThis 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.
- vonnik 11y agoI 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.