OpenAI’s new language model ChatGPT has been both praised and heavily criticized since its launch in late November 2022. Some people have gone as far as to say that it could take away our jobs and lead to rampant plagiarism in educational institutions. Before we get too worried, though, it’s worth investigating just how capable ChatGPT really is. To answer this question, OpenAI decided to put the model and its later counterpart, GPT-4, to the test by having them sit several professional and academic exams.
The results of the study, presented by OpenAI in a technical report, have been visualized in a chart. As it turns out, ChatGPT and GPT-4 are truly capable of human-level proficiency in many of these exams. For instance, GPT-4 was able to achieve a score better than 90 percent of test-takers in the Uniform Bar Exam, while ChatGPT only scored above 10 percent of all takers. Even though the more advanced model did noticeably better than its predecessor, there are still some exams that both models failed, such as AP English and competitive programming.
OpenAI is a research lab that focuses on artificial intelligence, robotics, and machine learning, among other things. It’s best known for its breakthroughs in natural language processing and automated dialogue systems, leading to the development of ChatGPT and the more advanced GPT-4. Through their research, OpenAI aims to advance the field of AI with a focus on safety and responsibility.
The person behind OpenAI is Sam Altman. He is one of the most prominent AI and entrepreneurship figures in the world. Along with Elon Musk, he co-founded OpenAI in late 2015 and still serves as the company’s CEO. Along with these accomplishments, he advocates for the ethical use of artificial intelligence and has developed innovative business models for startups.
