The use of artificial intelligence is becoming more widespread day by day, but it has its own problems. While image generation shows the number of fingers and so on which seems clear, but for text-based generation, it is more challenging to know which is right and which is wrong.
We already see AI generated results on Google producing inaccurate and wrong answers. Apart from that, now according to a Purdue University study, it was found that 52% of the answers generated by ChatGPT are related to programming, resulting in wrong answers.
For the purpose of this study, answers were compared for ChatGPT generation and also answers on Stack Overflow. It was seen that 52% produced wrong answers, 77% gave more words than normal human answers, and 78% gave inconsistent answers.
Further research also found that 35% of programmers chose the answer generated by ChatGPT, and around 39% of programmers did not find errors generated by ChatGPT. It is said to be one of the factors that make many may trust ChatGPT's answers more because of the more polite way of answering, compared to the answers on the forum by ordinary people who may sometimes be a bit rude.