⇧ [VIDÉO] You may also like this affiliate content (after ads)
If ChatGPT were a human, it would score between a B and a B —… The artificial intelligence developed by Open AI has demonstrated its ability to write convincing answers to the prestigious exam (MBA debriefing) offered by the Wharton School of Commerce.
“Not only are the answers correct, but the explanations are excellent as well,” says the research paper revealing this new test conducted with ChatGPT conversational AI. The goal was to require RNs to pass the standard exam that culminates in the MBA (Master of Business Administration), the highest level international training in business management.
The test was very convincing: “Chat GPT3 from OpenAI has demonstrated a remarkable ability to automate some of the skills of highly paid workers in general, and in particular, workers in positions requiring an MBA degree, including analysts, managers and consultants,” explains Wharton professor Christian Terwisz.
However, business school students have some consolation. ChatGPT certainly got a grade that would have allowed him to pass the exam, but he didn’t shine either. Christian Terwisz argues that AI would hardly have succeeded with a B or B-. Moreover, this score could only be obtained after some adjustments in the answers: the professor wanted to observe how the AI could improve when given hints in the face of an incorrect answer. “ChatGPT3 is remarkably good at modifying its responses after human signals,” he notes. “In other words, in cases where it initially failed to map the problem to the correct solution, Chat GPT3 was able to fix itself after receiving the appropriate hint from a human expert.”
“Massive” math errors
It should also be noted that the performance of ChatGPT is rather uneven depending on the domains covered by the questions offered. “As others before me have stated, Chat GPT3 sometimes makes startling mistakes in relatively simple calculations at the 6th grade math level. These mistakes can be massive,” says Christian Terwisz. Yesterday we mentioned in an article about the testing of ChatGPT and similar AI Claude by a specialized company. Similarly, the AI had difficulty solving certain mathematical problems.
In terms of analysis, ChatGPT also ran into its own limitations during this review: “The current version of Chat GPT cannot handle more complex process analysis questions, even if they are based on models. Pretty standard. This includes multi-product process flows and issues with stochastic effects such as demand volatility,” explains the professor. On the contrary, “Chat GPT3 excels at basic operations management and process analysis issues, including those based on case studies.” In any case, this speech is enough to raise some concerns… Both from the point of view of fraud and from the point of view of future competition in the labor market: disputes that are heard by many today, including in the artistic field.