But that’s the risk. ChatGPT gets its data from StackOverflow, no one goes to StackOverflow anymore so there’s no new data anymore.
One big advantage of StackOverflow is that some questions are supplemented years later with solutions that work in newer versions of frameworks. You will soon lose that and ChatGPT will only offer you solutions that worked or were incorrect at the time.
I often see people saying that ChatGPT gives them all the answers or writes complete code for them. My experience is that I have enough knowledge and come to ChatGPT only in exceptional cases, then 9/10 times I am disappointed, but I continue my search.
I’ve also found that ChatGPT is as bad as me at regular expression syntax. It only comes with a broken regular expression, including a broken Python example, although I put enough data in there with multiple examples of what the inputs are and what the output should be. In the end it wasn’t a complicated regular expression at all.
Since people really want to close the feedback loop and ChatGPT continues to train on what is ultimately a very large amount of code generated by ChatGPT itself, the quality will not improve very quickly.
“Coffee buff. Twitter fanatic. Tv practitioner. Social media advocate. Pop culture ninja.”
More Stories
Nivea responds after severe criticism of sunscreen: “We find it very disturbing.”
Enjoy pure nostalgia: discover the secret to juicy roast pork
OpenAI disbands the team that had to ensure the safety of its “super-intelligent” AI – IT Pro – News