Shownotes Transcript
Is ChatGPT Replace Developers?
—
Today's Amazon Deals - https://amzn.to/3FeoGyg)
—-
s ChatGPT Replace Developers?
Although ChatGPT is a useful tool for programmers who are competent enough to check and alter produced code, it is also completely lacking in intelligence. ChatGPT only works with text; it searches for information related to the request, such as code and explanations. It can “smoothly” combine text fragments, but it cannot reason logically or spot inconsistencies and absurdities in the result.
The only data used for training was from 2022. As a result of the model’s propensity to lag behind the most recent industry standards, I regularly produce out-of-date code.
Always keep in mind that this is a model that was trained using data, not a real software engineer, and that ChatGPT is the only source of human confirmation of the responses. As a result, different people will rate different answers as the “correct answer.”
Because a search engine’s data is more current than the model’s, recent tools, libraries, releases, etc. may need to be incorporated rather than just suggested and ChatGPT may fill in blanks with words or code that sound rational. It can be missing keys, libraries, or data structures. Alternatively it may just alter the names, the sequence, or anything else with each successive phase.
ChatGPT frequently has to be reminded to strengthen or harden code and is skewed towards straightforward instances without taking production considerations or edge circumstances into account and ChatGPT may produce redundant, incorrect, or even damaging ideas or code. It is unable to comprehend the logical interdependence of code and may offer generally sound counsel in inappropriate situations.
The input and output of ChatGPT are constrained in size and complexity. Although the specifics are private, you come upon them. For instance, if there isn’t enough “space,” the output may end mid-sentence or mid-code and ChatGPT may combine (in isolation) plausible procedures or libraries, but this would involve duplicating work. For instance, it might recommend compressing already-compressed material. Also, despite reminders that have previously fixed problems, it has a tendency to repeat them over time.