Educational institutions are at an inflection point, grappling with soaring costs, declining enrollments, and the rise in popularity of different educational alternatives. The latest threat to the education sector is ChatGPT, an AI-enabled chatbot developed by Silicon Valley startup OpenAI. The extensive language model (LLM) software was released on November 30th, 2022.
The use of ChatGPT for educational purposes has its own downsides. The software has been predicted to do away with higher education, code writing, journalism, and even the tech giant Google.
The End of Learning!
The introduction of ChatGPT in higher education realms has enabled it to write full-on student essays, solve mathematical problems, and even write elaborate codes to stir up the “artificial intelligence vs. education” debate.
Free and accessible tools like ChatGPT allow learners to complete homework assignments without much research and effort. However, it raises concerns about academic dishonesty, cheating, and the loss of learning ability. Instructors fear that learners will become lazy thinkers and eventually fail to develop lifelong research, critical thinking, and writing skills.
However, before concluding that artificial intelligence kills learning, it is wise to raise questions like why learners resort to such tools or other unethical practices (plagiarism or cheating) to write their essays.
One of the significant tech spending during the pandemic was proctoring solutions. Now, it is hard to come by a university that isn’t using plagiarism checker tools. While it may have provided short-term solutions to the wound left by the shift to online learning, it does not address the core problem: ChatGPT is not and will not be the only AI tool that will disrupt traditional learning methods. It is happening because the way we provide ai in education to students isn’t engaging, so they will always look for an easy way out.
The traditional ways of delivering lectures no longer engage students in the learning process. This disengagement had skyrocketed due to the pandemic. Due to the absence of physical contact in online classrooms, the constant interaction and connection between students and teachers are lost. Moreover, students have learned new disengagement behaviors such as muting themselves, turning off their cameras and skipping classes to watch the recorded lectures later. So, AI tools are bound to fill this gap and offer more engaging learning processes to students worldwide.
Though disruptions in conventional learning techniques offer an opportunity for higher education to adapt, the use of ChatGPT for learning purposes has many downsides.
Wrong Answers
As a language model, ChatGPT will invariably give incorrect responses. In some instances, it is incredibly incorrect. Andrew Ng, a computer scientist, thinks it’s “sometimes hilariously wrong.” However, because you can’t detect when ChatGPT is erroneous unless you already know the correct answer, as Arvind Narayanan points out on Twitter, ChatGPT occasionally seems to be highly confident in its responses even when they are incorrect (see below).
Training Data Limitations
ChatGPT is limited in its training data, like many other AI models. The training data biases and the lack of such data can reflect negatively on the results generated by ChatGPT.
The model generates results that are discriminatory in nature. In one such case, ChatGPT has demonstrated unfairness regarding minorities. Therefore, when the training data is biased, the generated results will be biased. This is a challenge with almost every AI tool, and we’ll have to make amendments to ensure fairness and minimize bias in technology.
Lack of Up-To-Date Information
ChatGPT is ill-equipped to perform tasks that require specialized knowledge, logic, and up-to-date information. Several observations demonstrate that it struggles with complicated mathematical calculations and arithmetic computations, suggesting that it hasn’t completely acquired logic.
Likewise, such models cannot be updated the same way we update knowledge bases by updating or replacing entities. The current model of ChatGPT has the training data derived from publicly available data before 2021, thus becoming incapable of generating accurate information promptly.
Final Thoughts!
Although ChatGPT is making waves in the tech world and has displayed suitability for many meaningful cases, the limitations mentioned above raise concerns about the usage of ChatGPT for educational purposes.
