Since its release to the public in November last year, OpenAI’s viral Artificial Intelligence-powered assistant ChatGPT has taken the online world by storm. Now, a growing number of professionals are feeling the heat as researchers lift the lid on the extent to which Artificial Intelligence tools could begin to replace or automate significant parts of their jobs.
While virtual chatbots aren’t exactly a groundbreaking new technology in and of themselves, the new capabilities present in OpenAI’s ChatGPT-3 have rapidly become one of the hottest topics of discussion on the internet.
ChatGPT gets its MBA
According to new research undertaken by Professor Christian Terwiesch at the University of Pennsylvania’s Wharton School, ChatGPT is now fully capable of passing a final exam for the school’s Master of Business Administration program.
Professor Terwiesch, who penned the research paper titled, ‘Would ChatGPT Get a Wharton MBA? A Prediction Based on Its Performance in the Operations Management Course,’ said that the bot scored somewhere between a B and a B minus on the exam.
In his research, Professor Terwiesch found that GPT-3 does its best work when its dealing with “basic operations management and process[ing] analysis questions.” According to the paper, the chatbot was able to provide correct answers for these sorts of questions along with solid explanations for how it arrived at said answers.
However, its worth noting that fears of an impending full-scale AI takeover may not be entirely warranted. In the paper’s summary, Terwiesch admitted that there are still a great many flaws in ChatGPT, with the bot making numerous mistakes on simple maths questions and struggled to handle more advanced lines of questioning in the MBA exam.
Professor Terwiesch’s findings come amidst a sea of new research that have looked at ChatGPT’s ability to achieve relative success in exams across a variety of different disciplines.
ChatGPT looks to medicine and law
Researchers have also tested out ChatGPT’s capabilities on the United States Medical Licensing Exam (USMLE). The USMLE is a three-part exam that all medical students in the US must sit if they want to practice medicine.
While researchers admitted they couldn’t leap to any conclusions due to their limited understanding of the new technology, they found that ChatGPT performed close to the passing threshold for the three exams.
A new paper titled ‘ChatGPT takes the Bar Exam‘ found that the while the AI-powered chatbot remains incapable of completing many of the longer more-complex sections within the Bar Exam, it performed well on the multiple-choice sections, scoring 50.3%, which is 50% higher than the baseline 25% guessing rate — which is roughly the score you’d get by picking answers at random.
ChatGPT also eclipsed the average passing rate for Torts and Evidence, two of the seven total sections of the Bar Exam Torts. Owing to these strong performances in fairly technical portions of the exam, the paper concludes, “we believe that these results strongly suggest that an LLM [large language model] will pass the MBE component of the Bar Exam in the near future.
Many commentators weren’t very impressed by ChatGPT’s performance, urging reporters to stop placing so much weight on the AI tool’s abilities. A criminal defence attorney from Vanguard Associates correctly pointed out that the chatbot had merely passed two multiple choice sections and that fears that lawyer’s jobs may soon be rendered obsolete, were unfounded.
Still, ChatGPT’s ability to perform on difficult examinations hasn’t gone unnoticed, with educators from around the world becoming increasingly concerned that AI tools could disrupt learning and assist students in cheating.
Even with all of this in mind, ChatGPT, despite being marketed as ‘Artificial Intelligence’ is still just a tool that aggregates data from the internet in a way that appears to be intelligent. While it can excel at some analytic areas on tests, through the aggregation of already-known data, it’s most likely still a long way from taking anyone’s jobs.