We’ve written about ChatGPT in the past, the conversational Artificial Intelligence (AI) language model that has permeated the national discussion. ChatGPT can produce grammatically and factually correct long-form essays on nearly any topic imaginable. ChatGPT wrote a passing AP English essay. It has passed a law exam at the University of Minnesota and scored 1020 on the SAT. It is very difficult to detect with traditional plagiarism checkers, because rarely are two responses the same.
Some students are leveraging ChatGPT for help with homework, long form essays and entrance exams.
But how smart is ChatGPT?
The graphic below compares the performance of ChatGPT 3.5 (the original version, released in November 2022) and ChatGPT 4.0, which was released on March 14, 2023, and is available for paid subscribers on a limited basis.
As the research shows, ChatGPT scores incredibly high – in the 99th percentile – on the Graduate Record Exam, which is required for entrance into most graduate schools. It also does well on the Bar Exam, the LSAT and many standardized AP tests.
Where it does less well is in English Literature and competitive computer programming.
What is notable is ChatGPT’s improvement between versions 3.5 and 4.0. On the Bar Exam alone it moved from the tenth percentile to the 90th.
In future versions of ChatGPT the line between man and machine is expected to blur due to the exponential rate of technology innovation (Moore’s law) and the increasing practicality of quantum computing.
But not yet. This article was written by a human. Or was it?
