Most computer science instructors will tell you that the Turing Test is more a theoretical or conceptual thought experiment than an actual exam that someone (or something!) can formally sit and receive a score on. A thread here on AI Stack Exchange confirms this.

Considering all of this, have there been any significant attempts to create a standardized form of a Turing Test that could be rolled out widely and used to evaluate various AI constructs? Obviously, none of these standardized testing systems could be considered The Only True Turing Test (TM), but perhaps they could have their place in research as a way to benchmark or categorize various algorithms or evaluate the work of students.

For example, I'm imagining hearing a graduate student muttering the following:

My AI construct passes the Johnson-Smith Turing Test 1992 and the Hernandez-Dorfer 2017, but it's still failing the Takahashi-2003 Advanced Elite. What am I doing wrong? Maybe if I tweak this routine here. [click]. Darn, still fails.

Either a fully automated test (e.g. just login and click to sit the exam) or a standardized system involving trained human judges (e.g. a la 21st century medical board exams) who apply standardized written rubrics would be acceptable as long as the criteria for passing are standardized rather than left to the judgment of untrained personnel or random passersby.



You must log in to answer this question.