OpenAI's GPT-4 Passes Bar Exam and Aces Standardized Tests
OpenAI unveils GPT-4, an advanced AI model that scores in the 90th percentile on major standardized tests like the bar exam and SAT without specific test preparation. The new system also processes images and writes in multiple coding languages.
OpenAI unveils GPT-4, its newest and most capable AI model, and claims the system achieves unprecedented stability and performance. The machine learning software scores in the 90th percentile or higher on the bar exam, the verbal GRE, and the reading and writing portions of the SAT. Remarkably, the system achieves these high marks without any specific training or studying for these standardized tests.
The new model outperforms the majority of human test-takers across a variety of academic subjects. GPT-4 earns a simulated 1410 on the SAT, passes the bar and GRE exams, and scores high enough on AP Art History, Biology, Calculus, and Chemistry exams to receive college credit. The AI even excels at sommelier tests, suggesting potential future applications in writing wine bottle descriptions.
Building on the foundation of the widely discussed GPT-3, this latest generative pre-trained transformer introduces powerful new capabilities. GPT-4 writes in multiple coding languages, generates narrative scripts, provides step-by-step instructions for complex questions, and interacts with images. As OpenAI releases this highly advanced model to the public, discussions about AI's impact on disinformation, education, and the workforce continue to grow.