Uh hi everyone, my name is Assistant Professor in the Department of Computer Science and Engineering. Today, I will present our research project on the coding performance of GBT (Generative Pre-trained Transformer). The project aims to study the performance of GBT in coding tasks and analyze various aspects related to its usage.
The field of Artificial General Intelligence (AGI) is rapidly emerging, and GBT is one of the prominent technologies in this field. GBT has the capability to generate code based on given prompts or inputs. For example, it can generate code for sorting algorithms like the Bubble Sort algorithm, which is a fundamental algorithm in computer science.
In our research project, we have six research questions that we aim to answer. Firstly, we want to identify the programming languages supported by GBT and determine which programming languages are most widely used by people. Secondly, we analyze the scenarios, tasks, and purposes for which people use GBT for programming. Thirdly, we investigate how people perceive the code generated by GBT and analyze their sentiments towards it. Fourthly, we examine the quality of the code generated by GBT and evaluate its performance using industry-standard metrics. Lastly, we explore the ethical implications of using GBT for code generation.
To conduct our research, we adopted a crowd-sourced methodology. Instead of testing GBT ourselves, we collected data from online platforms like Twitter and Reddit, where people discussed GBT’s coding performance. We used natural language processing and computer vision techniques to analyze the data and extract insights.
Our findings show that Python is the most popular programming language supported by GBT, followed by JavaScript and R. We also identified three main types of tasks for which people use GBT: programming interviews, school assignments, and coding, writing, and debugging. Additionally, we observed that people often discuss GBT in relation to AI, machine learning, and data science.
When it comes to how people perceive the code generated by GBT, our sentiment analysis revealed that the dominant emotion is fear. People express concerns about being replaced by GBT in the future. However, we also found that the code generated by GBT is of good quality, with only a small percentage of syntax errors.
In terms of ethical implications, we found that GBT can generate code that raises ethical concerns, such as predicting gender based on given prompts. However, GBT also has the capability to refuse certain tasks that are considered unethical.
In conclusion, our research project sheds light on the coding performance of GBT and its various aspects. GBT is widely used in coding tasks and has both positive and negative perceptions among users. Further exploration of the ethical implications of AI-generated code is necessary for the future development and deployment of GBT.
