ChatGPT
(Updated May 26, 2023)
(Updated May 26, 2023)
ChatGPT (chatbot) is constantly being assessed regarding its test performance effectiveness against all TAC knowledge test versions.
Currently, ChatGPT is minimally effective (15%-22% of questions answered correctly); a two-percentage range reduction from the previous trial result.
We are effectively eliminating questions that work well with AI by replacing those questions, formats and communique types with content proven difficult or implausible for present day AI to address correctly. View NorthStar's FIC below.
Focused Information Criteria (FIC)
-Verbal vs. Nonverbal.
-Questions/Formats developed post 2022.
-AI hallucination.
In an effort to reduce AI test effectiveness to a sub-five percentage range, a new criterion is examined on a monthly basis.
Currently, ChatGPT is minimally effective (15%-22% of questions answered correctly); a two-percentage range reduction from the previous trial result.
We are effectively eliminating questions that work well with AI by replacing those questions, formats and communique types with content proven difficult or implausible for present day AI to address correctly. View NorthStar's FIC below.
Focused Information Criteria (FIC)
-Verbal vs. Nonverbal.
-Questions/Formats developed post 2022.
-AI hallucination.
In an effort to reduce AI test effectiveness to a sub-five percentage range, a new criterion is examined on a monthly basis.
