Is ChatGPT Harmful or Helpful to Business Professionals? Part 2 The Bad

B. The Bad

As ChatGPT was trained by both human and machine trainers prior to its public release, given the immense amount of misinformation on the web, it can provide erroneous answers. For instance, when asked a simple question about a book, it may offer a mix of correct and incorrect content - and different wrong answers each time the response is generated. Although ChatGPT knows its limitations and offers a disclaimer warning to the users about potential inaccuracies, the potential that users could become overly reliant on the chatbot’s answers without verifying their accuracy. Indeed, one of the major concerns is that ChatGPT can provide a partially correct answer with a very authoritative-sounding explanation or reasoning behind it. Thus, to rely on the chatbot to provide a completely accurate answer each time, or to even believe that it could do so is a significant issue that users must be aware of.


Another drawback of ChatGPT is that its knowledge doesn't extend beyond 2021, so it cannot comment on recent events such as the death of Queen Elizabeth II. As tax rules change frequently, ChatGPT in its current version cannot provide accurate and up-to-date answers. For example, in Screenshot 15, the author asked the chatbot a question about the deductibility of business expenses for meals and entertainment. ChatGPT replied with a logical and authoritative explanation as to why the entire amounts were non-deductible. However, it was only partially correct as in 2021 and 2022 tax rules changed so that the qualified meal expenses at a restaurant are fully deductible for business purposes.


Beyond taxation concerns, ChatGPT can generate potentially dangerous information such as creating the basic blueprint to construct a timed bomb or how to generate poisonous gas from various elements. For those sensitive to political correctness, some answers given by ChatGPT could be perceived as insensitive. Thus, although disclaimers accompany the chatbot’s  instructions or answers, they could potentially be used to aid those who have malicious intent in the dissemination of such information or use the answers as proof of correctness despite the fact that ChatGPT has no authority over all subject matters. 

In addition to potential harm to humans and society, ChatGPT could potentially be used to cheat in an educational setting and undermine students’ learning process. For example, when tested more than 25,000 accounting and tax assessment questions in various formats (e.g., open ended short answers, multiple-choice questions, and essays), researchers found that ChatGPT slightly outperformed the average students by about sixteen percent. As this advanced chatbot is publicly available for free to users throughout the globe, students could be tempted to use it to cheat on assessments such as quizzes and exams or by having ChatGPT write essays for them and pass it as their own writing. While there have been some tools developed to detect whether texts are AI-generated, the danger of students using it to bypass actual deep learning could limit their attainment of true knowledge in the field and thus potentially cause harm to the next generation of tax accountants and practitioners


Next week, I will continue with part 3 of the ChatGPT's impact. If you enjoy my posts, please click on the blue follow button by my profile picture and comment.

Jason Chen, PhD

Comments

Popular posts from this blog