GPT-4, the latest version of the artificial intelligence (AI) chatbot, ChatGPT, can pass high school and law school exams with scores in the 90th percentile and has new processing capabilities that were not possible with the previous version.
Data from the GPT-4 tests were shared On March 14, its creator, OpenAI, revealed that in addition to handling “much more detailed instructions” more creatively and reliably, it can also convert image, audio and video input into text.
“He passes the mock bar exam with scores around the top 10% of test takers,” OpenAI added. “In contrast, the GPT-3.5 score was around the bottom 10%.”
The numbers show that GPT-4 scored 163 in the 88th percentile on the LSAT, the test that college students in the United States must take to be admitted to law school.
A GPT4 score would put her in a good position to get into a top 20 law school and she is just a few grades short. reported score needed for admission to prestigious schools such as Harvard, Stanford, Princeton or Yale.
The previous version of ChatGPT scored only 149 on the LSAT, putting it in the bottom 40%.
GPT-4 also scored 298 out of 400 on the Uniform Bar Exam — a test taken by recently graduated law students that allows them to practice law in any US jurisdiction.
The old version of ChatGPT struggled in this test, finishing in the bottom 10% with a score of 213 out of 400.
On the SAT Evidence-Based Reading & Writing and SAT Math exams taken by US high school students to measure their college readiness, the GPT-4 scored in the 93rd and 89th percentiles, respectively.
GPT-4 also excelled in the “hard” sciences, recording well above average percentile scores in AP Biology (85-100%), Chemistry (71-88%), and Physics 2 (66-84%).
However, his AP Calculus scores were fairly average, ranking between the 43rd and 59th percentiles.
Another area where the GPT-4 was lacking was in the English literature exams, where scores in the 8th to 44th percentiles went into two separate tests.
OpenAI said that GPT-4 and GPT-3.5 passed these tests from the 2022-2023 field trials, and that the language processing tools did not undergo “any specific training”:
“We did not undergo any specific training for these tests. A minority of the problems in the tests were seen by the model during training, but we believe the results are representative.”
The results also caused fear in the Twitter community.
Related: How will ChatGPT impact the Web3 space? The industry responds
Nick Almond, founder of FactoryDAO he said to his 14,300 Twitter followers on March 14 that GPT4 will “frighten people” and “crash” the global education system.
Evaluation theory has been a big part of my life for several years. Many years ago I alluded to this day. I literally sounded like a resident at the time.
But… this actually means that from this point on, everything but the uncontrolled rating is over.
— drnick ️² (@DrNickA) March 14, 2023
Former Coinbase CEO Conor Grogan said he put a live Ethereum smart contract into GPT-4 and immediately pointed out several “security vulnerabilities” and outlined how the code could be abused:
I put a live Ethereum contract into GPT-4.
In no time, he pointed out a number of security weaknesses and pointed to superficial areas where the contract could be abused. He then verified the specific way I could use the contract pic.twitter.com/its5puakUW
— Conor (@jconorgrogan) March 14, 2023
Previous smart contract audits on ChatGPT he found that his first version was also reasonably capable of detecting errors in the code.
Rowan Cheung, founder of AI newsletter “The Rundown”, shared a video of GPT transcribing a hand-drawn fake website on a piece of paper into code.
I just watched GPT-4 turn a hand-drawn sketch into a working website.
That’s crazy. pic.twitter.com/P5nSjrk7Wn
— Rowan Cheung (@rowancheung) March 14, 2023
