Ever since artificial intelligence (AI) chatbot ChatGPT made its debut in November. It has become the topic of every conversation on the internet thanks to its chat skills. People are asking different things to ChatGPT in a playful and light manner to test the capabilities and power of the chatbot.
The tool has recently passed some prominent tests, including the US medical licensing exam. A Wharton Business School exam for the MBA program’s operations management course final exam. And four University of Minnesota Law School exams in constitutional law.
Now, reacting to ChatGPT passing the US medical licensing exam. Twitter CEO Elon Musk has taken a dig at the artificial intelligence chatbot. Mr Musk responded to a tweet, which touted the chatbot’s capabilities, saying, “I’m sure everything will be fine.”
Notably, AI research firm OpenAI, in which Elon Musk and Microsoft have invested capital, developed the chatbot ChatGPT. Anyone using the AI Tools website can ask the chatbot a question about anything and get a quick, detailed response in paragraph form.
Over the past few weeks, ChatGPT has demonstrated what it’s capable of. The tool has written quick and complex essays, drafted marketing pitches, created poems and jokes. And even created a speech for a US congressman.
However, there are also fears that AI could take some human jobs.
But in its latest achievement, the AI tool has even passed the US medical licensing exam. In a pre-print study, researchers explored the upper limits of chatbot capabilities, according to ABC News. They say ChatGPT has achieved over 50 percent on one of the toughest standardized tests around: the US Medical Licensing Exam (USMLE).
Also Read | The Emirates plane flew for 13 hours, landing at the same place it took off from
To test the program’s viability, researchers administered it to a mock, shortened version of the USMLE. Which is required for any doctor to be licensed to practice medicine in the United States. The researchers fed questions from previous experiments into the AI tool. And the answers, ranging from open-ended written responses to multiple-choice, were scored by two clinical judges. They also ensured that the answers to these questions were not already in the dataset when the chatbot was trained.
The team noted that even though ChatGPT had not already seen the answers. It performed at or near the passing threshold for the test without any special training or reinforcement. The tool scored over 50 percent on all tests and reached the USMLE pass threshold of about 60 percent. “Therefore, ChatGPT is now comfortably within the passing range,” the paper concludes.