Tech

New Version Of ChatGPT Crushes LSAT, SAT, GRE And AP Exams

(Photo by Jason Redmond / AFP) (Photo by JASON REDMOND/AFP via Getty Images)

Font Size:

The latest iteration of the ChatGPT artificial intelligence has reportedly scored well enough on multiple standardized exams to gain admission to selective higher education institutions.

GPT-4, the newest version of the ChatGPT technology, made waves on social media when several Twitter users noted that the AI was able to score very highly on exams including the SAT, LSAT, GRE, Advanced Placement tests and the bar exam. OpenAI announced the update of the technology Tuesday, claiming that it holds human-level capabilities on several professional benchmarks. The company cautioned, however, that the chatbot still comes up short in some areas and cannot fully replicate human performance in all fields.

“It is still flawed, still limited, and it still seems more impressive on first use than it does after you spend more time with it,” OpenAI CEO Sam Altman tweeted.

The performance upgrade from GPT-3.5, the previous iteration of the artificial technology, are significant. OpenAI notes that GPT-4 scores in the top decile on the bar exam, whereas GPT-3.5 scored in the bottom 10 percent. The company also claims that the new version of the technology is more capable of handling complex tasks. GPT-4 managed to score in the 90th percentile of the SAT, the 99th percentile of the verbal GRE, and a 5 (the highest score) on the AP Economics and AP Biology exams.

The AI’s impressive performance on exams raised eyebrows online.

“the big thing that gpt4 makes obvious is that the entire field has moved away from esoteric NLP benchmarks to benchmarking against things that humans actually do,” Will Manidis, CEO of ScienceIO, wrote on Twitter.

Meanwhile, journalist Matthew Yglesias joked that GPT-4’s test results should be a point of pride for English majors.

“English majors get the last laugh as GPT-4 crushes every exam except AP English Language and AP English Lit,” Yglesias tweeted.