Chat-GPT-3 is one of the large-scale language model’s recently attracting attention for its potential to discuss and generate code (Tian et al., 2023). Effectiveness and accuracy varies on the specific task, yet GPT-3’s advantages comes down to its ability to process additional information from the user, increasing its success rate (Surameery & Shakor, 2023). GPT-3 can assist developers in identifying bugs and possible solutions, yet the quality of the output varies significantly and is often dependent on human collaboration to achieve the best result.

There was an opportunity to utilise GPT-3 in the development of the JS code, as there was a lack of knowledge with the language. It was found successful in explaining code, and somewhat accurate in identifying and fixing bugs, yet unnecessarily complicated in implementation of code to visualise elements or execution of functions. The model struggles to provide solution to new and un-seen problems. The code was modified with the help of other resources and human intervention, yet remains mainly GPT-3. As can be seen by the implemented JS code, the code is somewhat messy and uneccessarily long or complicated, however still works. Through several prompts worded in different ways, providing possible solutions or errors, as well as being specific to exactly what was to be implemented or in need of modification, it resulted in the submitted outcome. The code was kept mainly AI generated to show the result of using GPT-3 as a co-developer by not only fixing bugs but developing code.

LLM’s performance remains limited but is contributing to future automatic software reparation with minimal human intervention (Tian et al., 2023). To ensure the best outcome, GPT-3 should be used as a debugging tool in conjunction with the strength of other debugging tools, providing developers with a better understanding of the code by identifying and fixing bugs more efficiently (Surameery & Shakor, 2023). It is currently inadequate in more complex code generating.

Bibliography

Surameery, N.M. and Shakor, M.Y. (2023) ‘Use chat GPT to solve programming bugs’, International Journal of Information technology and Computer Engineering, (31), pp. 17–22. doi:10.55529/ijitc.31.17.22.

Tian, H. et al. (2023) Is ChatGPT the Ultimate Programming Assistant — How far is it?, pp. 1–12. doi:10.48550/arXiv.2304.11938.