As if high school students weren’t stressed enough about standardized tests and college applications, a robot has entered the chat.<\/p>\n\n\n\n
Earlier this month, OpenAI announced GPT-4, its latest large language model, which improves on predecessor ChatGPT’s reasoning capabilities and demonstrates “human-level performance” on a variety of standardized tests, including the bar exam, the LSAT, the GRE, a slew of AP tests, and even sommelier exams.<\/p>\n\n\n\n
According to OpenAI’s technical report for GPT-4<\/a>, the model’s scores<\/a> were mostly excellent and occasionally exceptional, with a few blunders, such as its abysmal score on AP English Literature and Composition. (The robots may take our jobs, but they will never take our literary criticism.)<\/p>\n\n\n\n Notably, GPT-4 performed well on its SATs, the standardized test widely used for college admissions in the U.S. It scored 710 out of 800 on the evidence-based reading and writing (ERW) portion of the SAT and 700 out of 800 on the math portion for a total score of 1410 out of 1600.<\/p>\n\n\n\n Comparatively, the average SAT score in 2022 was 1050, according to the College Board<\/a>, with a score of 529 on ERW and 521 on math. GPT-4’s SAT score of 1410 puts it in the 94th percentile<\/a> of students who took the SAT nationally.<\/p>\n\n\n\n Build the confidence and practical skills that employers are looking for with Forage’s free job simulations.<\/p>\n\n\n\n Of course, college admission requirements \u2014 including standardized test scores \u2014 vary by institution. With the SATs continuing to serve as a hot (and polarizing) topic, we were curious as to how GPT-4 might fare this college application season based on its performance.<\/p>\n\n\n\nShowcase new skills<\/h2>\n\n\n\n
Assessing the Model\u2019s Odds of Admission: Methodology<\/h2>\n\n\n\n