ChatGPT version 4 successfully passes SATs being in the top 10% of the cohort

Updated: 15 March 2023

Disclosure

Don’t invest unless prepared to lose all the money you invest. This is a high-risk investment, you shouldn’t expect to be protected if something goes wrong.

Join Our Telegram channel to stay up to date on breaking news coverage

ChatGPT v4, also known as GPT-4, is the latest version of OpenAI’s revolutionary chatbot that has recently captivated the tech industry. The new version reportedly has the ability to successfully pass high school and law school tests, completing them with scores that allowed it to rank among the top 10% of participants.

The developers also improved the bot’s processing capabilities, making them much more advanced than what the past versions had to offer.

ChatGPT v4 test results are impressive in nearly all areas

OpenAI released the figures of GPT-4 test scores yesterday, March 14th. The details also revealed that the bot is now capable of converting images, audio, and even video inputs into text. The tests have revealed that the bot can also easily handle more nuanced instructions, showing increased reliability and creativity.

OpenAI’s report noted that

It passes a simulated bar exam with a score around the top 10% of test takers. In contrast, GPT-3.5’s score was around the bottom 10%.

After being tasked with the completion of the LSAT exam, the bot achieved a score of 163, landing in the 88th percentile on the exam. The LSAT exam is a test that college students are required to pass in the US in order to be admitted into law school.

With a score like this, the bot would technically be accepted into any of the top 20 law schools in America. If it scored only a few points higher, it would also qualify to enter some of the most prestigious schools, like Stanford, Harvard, Princeton, and Yale. This means that the GPT-4’s results significantly exceed the prior version, which only scored 149 when exposed to the same LSAT test. With a score like that, it landed in the bottom 40%.

OpenAI was also curious to see how well GPT-4 would handle the Uniform Bar Exam, discovering that it had scored 298 out of 400. Uniform Bar Exam is a test taken by law school graduates, and passing it grants them permission to practice law in any jurisdiction in the United States.

When taking the same test, the past version of ChatGPT scored only 213 out of 400. ChatGPT v4 also tried out SAT Evidence-based Reading & Writing, scoring 93%, and SAT Math exam, scoring 89%. Testing it against the “hard” sciences, the developers said that it scored quite well, with above-average percentile scores in Physics 2 (66-84%), Chemistry (71-88%), and AP Biology (88-100%)

Surprisingly, the bot struggled when it came to AP Calculus, with a rank of 43-59%. However, the bot struggled the most when it came to English literature exams, with scores in the 8-44%. OpenAI claims that they did no special training of the bot to prepare it for the exams.

‘GPT-4 will scare people,’ according to the Twitter community

The Twitter community had some strong feelings on the matter, with some convinced that GPT-4 will scare people and potentially lead to the collapse of the global education system.

Assessment theory was a big chunk of my life for several years. I was banging on about this day coming many years ago. I literally sounded like the resident crank at the time.

But… really this means that anything but invigilated assessment is over from this point on.

— drnick (@DrNickA) March 14, 2023

Another interesting tweet came from Conor Grogan, the former director of Coinbase, who said that he dumped a live ETH smart contract into the new version of the bot, and it managed to instantly identify multiple security vulnerabilities.

I dumped a live Ethereum contract into GPT-4.

In an instant, it highlighted a number of security vulnerabilities and pointed out surface areas where the contract could be exploited. It then verified a specific way I could exploit the contract pic.twitter.com/its5puakUW

— Conor (@jconorgrogan) March 14, 2023