This post will discuss how ChatGPT-4 smashed the Base SAS Certification with a score of 93.4% in only half the time allotted for the exam.
GPT-4 is a huge improvement over the prior version GPT-3.5. OpenAI Team claims that as compared to the previous GPT-3.5 models, GPT-4 makes much fewer mistakes in generating factual statements. On factuality evaluations that are designed to challenge the models, GPT-4 performs 19 percentage points better than GPT-3.5. Microsoft’s new Bing Chat also solves the limitation of GPT-4 training data only upto September 2021. In other words, Bing Chat can answer to the questions related to recent events.