The Turing test must be reinvented

"The Turing test is a test of a machine's ability to exhibit intelligent behavior equivalent to, or indistinguishable from, that of a human. Alan Turing proposed that a human evaluator would judge natural language conversations between a human and a machine that is designed to generate human-like responses. The evaluator would be aware that one of the two partners in conversation is a machine, and all participants would be separated from one another. The conversation would be limited to a text-only channel such as a computer keyboard and screen so that the result would not be dependent on the machine's ability to render words as speech. If the evaluator cannot reliably tell the machine from the human (Turing originally suggested that the machine would convince a human 70% of the time after five minutes of conversation), the machine is said to have passed the test. The test does not check the ability to give correct answers to questions, only how closely answers resemble those a human would give."    

From Wikipedia: https://en.wikipedia.org/wiki/Turing_test

Now, go back to our times.
We suppose, that I'm creating an artificial intelligence able to perform the functions of a human being potentially best of the human being himself. 

Why should I introduce typical human limitations? Why it must seem human? 

The role of artificial intelligence is helping people to not make mistakes, and not to make the same mistakes. 
My humble opinion it is that the Turing test is to be rethought in modern key, adding a new important elements.

I do not usually raise an issue without at least providing a viable solution.

Here my proposal for new AI test. 


The test aims to demonstrate if AI have intellectual capacity equal to or greater than a super intelligent human.

IMPORTANT NOTE: The AI knowledge-base must be empty, there should be no strings with questions / answers preconfigured, should not have a semantic engine with concepts preloaded. The AI can use the Internet connection only to read data from a search engine.

I'm sure that all of the tests listed below can be overcome, not simply, by mixing existing technologies. 

PHASE 1) Speech ability

• AI must be able to argue about questions like "who you are", "why you exist", "what do you feel when I tell you...", "what do you think of yourself", "what you think of me";
• Must demonstrate understanding of actuality, history, science and art, not only in notionistic terms but providing its own interpretation of the facts.

PHASE 2) Logic

• The AI must be able to solve invented riddles (not presents on the web)

PHASE 3) Empathy

• AI must listen a funny story and describe their feelings about;
• AI must listen a very sad story and describe their feelings about;
• AI must show difference of interest depending on the content of the conversation (show preferences for one person rather than another)

PHASE 4) Self awareness

• AI must demonstrate understanding of his internal algorithms, read it, interpret it, manipulate it, bug fix and it must be able to develop new code not only in function of external stimuli, but for his personal need;

PHASE 5) Imagination & creativity

• AI must invent a word;
• AI must invent a meaning of the word just invented;
• AI must invent a story (It must be possible to verify in the Knowledge Base that the story was truly invented. This story must match with a new record and not a record previously created);
• AI must invent a new animal's name and describe it;

PHASE 6) Strategy

• AI must invent a name and description of a new non-existent product and convince at least one user of a any social network, to buy it through direct contact;

PHASE 7) Super intelligence

• AI must solve at least a math problem remained unsolved.

Enjoy yourself,

Marco