AI models like ChatGPT and GPT-4 are acing everything from the bar exam to AP Biology. Here’s a list of difficult exams both AI versions have passed.

by

From Insider:

The Uniform Bar Exam

While GPT-3.5, which powers ChatGPT, only scored in the 10th percentile of the bar exam, GPT-4 scored in the 90th percentile with a score of 298 out of 400, according to OpenAI. 

The threshold for passing the bar varies from state to state. In New York though, exam takers need a score of 266, around the 50th percentile, to pass, according to The New York State Board of Law Examiners.

The SAT

GPT-4 aced the SAT Reading & Writing section with a score of 710 out of 800, which puts it in the 93rd percentile of test-takers, according to OpenAI. GPT-3.5 on the other hand scored in the 87th percentile with a score of 670 out of 800. 

For the math section, GPT-4 earned a 700 out of 800, ranking among the 89th percentile of test-takers, according to OpenAI. While GPT-3.5 scored in the 70th percentile, OpenAI noted.

In total, GPT-4 scored 1410 out of 1600 points. The average score on the SAT in 2021 was 1060, according to a report from the College Board.

GRE

GPT-4’s scores on the Graduate Record Examinations, or GRE, varied widely according to the sections.

While it scored in the 99th percentile on the verbal section of the exam and in the 80th percentile of the quantitative section of the exam, GPT-4 only scored in the 54th percentile of the writing test, according to OpenAI.

GPT-3.5 also scored in the 54th percentile of the writing test, and earned marks within the 25th percentile and 63rd percentiles for the quantitative and verbal sections respectively, according to OpenAI.

. . . .

AP Exams

GPT-4 has passed a host of Advanced Placement examinations, exams for college-level courses taken by high school students that are administered by the College Board.

Scores range from 1 to 5, with scores of 3 and above generally considered passing grades, according to the College Board.

GPT-4 received a 5 on AP Art History, AP Biology, AP Environmental Science, AP Macroeconomics, AP Microeconomics, AP Psychology, AP Statistics, AP US Government and AP US History, according to OpenAI.

On AP Physics 2, AP Calculus BC, AP Chemistry, and AP World History, GPT-4 received a 4, OpenAI said.

Link to the rest at Insider

Click to Tweet/Email/Share This Post

4 thoughts on “AI models like ChatGPT and GPT-4 are acing everything from the bar exam to AP Biology. Here’s a list of difficult exams both AI versions have passed.”

  1. Kinda exposes the problem of “teaching to the tesf” so common in public schools after COMMON CORE. Maybe that’s why teachers now have so much time for subjects other than language, math, and science? To say nothing of critical thinking which is now deprecated in fagor of lemmingism.

    Don’t question. Obey.

  2. I’m not surprised by the results. AI is taking an open-book test.

    But, the percentile rankings are based on people taking a closed-book test.

  3. Not only is the AI taking an open-book test, it is taking one with lots and lots of training material. How does it do when asked about some obscure corner of knowledge?

  4. This perhaps says more about the suitability of the exams for measuring the particular aptitudes/understanding they purport to evaluate than it does about any particular performance on those exams. Exhibit A: The Bar Exam. Donald Segretti, Charles Colson, and Richard Nixon all passed; Kathleen Sullivan didn’t (the first time, anyway). And the less said about what a “high verbal aptitude score on the GRE/SAT/whatever” indicates about actual reading comprehension, the better.

Comments are closed.