Profile

Cover photo
Chao Jia
18 followers|16,269 views
AboutPostsPhotosVideos

Stream

Chao Jia

Shared publicly  - 
 
I won candy on the #googlebirthday doodle! Score: 120 (Happy birthday, Google)
1
Add a comment...

Chao Jia

Shared publicly  - 
 
Finally I understand why the posts you +1 in Google plus don't show up in +1's list...They want you to share it...
 
From someone who isn't on here, yet.
1
Add a comment...

Chao Jia

Shared publicly  - 
 
听哥德堡变奏曲,又被感动了,古尔德1981年版本
 ·  Translate
1
Add a comment...

Chao Jia

Shared publicly  - 
 
Apple really sucks.......
1
Add a comment...
Have him in circles
18 people
Zhifang Chen's profile photo
Ramon Hu's profile photo

Chao Jia

Shared publicly  - 
 
 
The widely popular notion of the Turing Test being employed to verify "true" AI is off-base

Turing test involves a conversation between a human evaluator and an AI. If the evaluator cannot tell whether they're speaking with another human or a machine, the AI passes the test. If this seems like a reasonable validation of AI, why has the Turing test fallen out of favor with AI scientists?

First, the winners of the Loebner Prize (turing test contest) turn out to be chat-bots that misdirect the evaluator. These chat-programs are cleverly written to misdirect, but are not actually intelligent. Google Now and Apple Siri are not going to win the Loebner Prize, but are, in fact, much more intelligent entities. 

Second, it is extremely unlikely that any computational agent will ever have the same experience growing up as a human. Thus, it will be extremely easy to tell whether you are speaking to a human or a machine. Imagine a conversation, where a human evaluator asks, "how's the weather where you are?" and one machine answers, "great! sunny and beautiful here :)" and another machine answers, "i can't imagine why humans are so fascinated with minor atmospheric changes." The first of these will win the Turing test, whereas the second may be the more intelligent of the two.

Machines will likely grow up with different sensory, motor, and hormonal systems than humans, and, even if we could make humanoid machines perfectly replicating human sensory/emotional capabilities, machines will always be treated different from humans growing up, they will look different, they will not have biological parents there for them 24x7, will not have human toddlers to interact with during their toddler stage, nor will they even ever get to be toddler-size, growing through the same physical changes as human beings. "You know how you come back to the place where you grew up and it seems smaller?" type of a conversation will be lost on a robot. 

As of today, a better Turing test is one where humans and machines are evaluated in simpler environments, removing the potential to discriminate based on cultural recognition. For example, the evaluator may play a game of poker or FIFA soccer against an AI, and judge whether their opponent was a human or a machine. If the AI can pass for a human across multiple task environments, it is deemed to have passed this modified version of the Turing test. 
1
Add a comment...

Chao Jia

Shared publicly  - 
1
Add a comment...

Chao Jia

Shared publicly  - 
 
This guy really has a huge crush on the history of United States :-)
1
Add a comment...

Chao Jia

Shared publicly  - 
 
Great people talk about IDEAS.
Average people talk about THINGS.
Small people talk about OTHER PEOPLE.
1
Add a comment...
People
Have him in circles
18 people
Zhifang Chen's profile photo
Ramon Hu's profile photo
Links
YouTube
Contributor to
Basic Information
Gender
Male