The newly released version of OpenAI’s chatbot, GPT-4, has outperformed its predecessor, GPT-3.5, with flying colors. OpenAI announced that GPT-4 has the ability to pass high school and law school exams with scores in the 90th percentile, as well as find exploits in smart contracts written in Solidity.
The exponential growth of artificial intelligence (AI)
ChatGPT, the revolutionary AI chatbot, continues to make headlines and transform how we look at work. The newest version, GPT-4, can process much more complex and nuanced instructions in a more creative and reliable manner.
OpenAI released the figures for GPT-4’s test scores on March 14, revealing that the chatbot can now convert image, audio, and video inputs into text, making it a more versatile tool.
Becoming a brilliant academic
The simulated bar exam produced the most impressive results for GPT-4, achieving a score within the top 10% of test takers, a stark contrast to GPT-3.5, which scored in the bottom 10%. GPT-4 also scored 163 in the 88th percentile on the LSAT exam, placing it in an advantageous position for admission to a top 20 law school. In comparison, the previous version of ChatGPT, only managed a score of 149 on the LSAT, ranking in the bottom 40%.
In addition, GPT-4 performed well on the Uniform Bar Exam, scoring 298 out of 400, while the previous version of ChatGPT struggled, finishing in the bottom 10% with a score of only 213 out of 400. The Uniform Bar Exam is taken by law graduates to acquire the license to practice law in any jurisdiction in the United States.
GPT-4 demonstrated proficiency in the “hard” sciences, achieving scores well above the average percentile in AP Biology, Chemistry, and Physics 2. However, its AP Calculus score was only average, placing in the 43rd to 59th percentile. On the other hand, GPT-4’s performance was less impressive in English literature exams, scoring in the 8th to 44th percentile across two separate tests.
OpenAI has officially stated that GPT-4 and GPT-3.5 took the 2022-2023 practice exams without any specific training, yet GPT-4 managed to outperform its predecessor by a significant margin. These results have sparked concern among some Twitter users who fear that GPT-4’s capabilities could be intimidating enough to cause a potential breakdown in the global education system.
ChatGPT helps blockchain devs write and audit smart contracts
GPT-4 has demonstrated its skill in detecting security flaws in Ethereum smart contracts, along with its academic achievements. Conor Grogan, a former director at Coinbase, incorporated a functioning Ethereum smart contract into GPT-4, which immediately identified numerous “security vulnerabilities” and described ways in which the code could be exploited.
ChatGPT has already shown its aptitude for detecting code bugs during previous smart contract audits. However, this functionality was way less advanced.
The founder of the AI newsletter The Rundown, Rowan Cheung, recently shared a video demonstrating GPT-4’s impressive ability to transcribe a hand-drawn mockup of a website onto paper and convert it into functioning code. This video is a clear testament to the versatility of GPT-4 and the vast potential it has in numerous industries.
The release date and usage plans for GPT-4 remain undisclosed by OpenAI. However, this latest update has the potential to make a significant impact in various fields, such as education, law, and cybersecurity. The precise manner in which the chatbot will be incorporated into these industries remains to be seen, but the scope of its versatility and expertise is clear.
