The ChatGPT Artificial Intelligence model continues to make history in the technological world. Now they put it to the test against the last PAES for Reading Comprehension, and the one for History and Social Sciences. In them, the AI managed to get most of the questions right, and if it were an applicant, it would have positioned itself within the 1% of the best scores obtained in the exam.
The PAES is the instrument used in Chile so that students about to graduate from college and from previous generations can access different universities. This achievement is an important milestone in the field of AI and shows the potential that this technology has to face increasingly complex challenges.
For performance tests EvoAcademy used 2 versions of ChatGpt. First, when using the free version 3.5 to answer questions of Reading Comprehension and History and Social Sciences of the last PAES Test, he obtained a success rate of more than 75%. Later it was the turn of GPT4, which is the paid version, and here the accuracy exceeded 93%, failing only one question that resulted in 917 points in the exam. there is a difference of almost 200 PAES points between both versions.
As mentioned by EvoAcademy, “This exercise should not be considered a finished and rigorous study. However, these preliminary results invite us to reflect and ask ourselves about the use of AI in education.”
Editorial: Gaming / Facebook / Twitter / Coverage / Instagram / Discord