I’m working on a 200-page Word document with feedback from over 1200 people. All I need are answers to 2 or 3 specific questions, but every time I ask ChatGPT, the results change.
Is this normal or am I doing something wrong?
Welcome to the forum! Here are some tips to make the most of your posts:
- Add as many details as you can to your questions. It helps everyone understand the problem better.
- If you’re asking about a common topic, try searching the forum first to see if it’s already been discussed.
- Healthy debates about AI pros and cons are great, but remember to keep it respectful.
- If you’re making a claim, share some links or examples to back it up.
Feel free to reach out to the mods if you have any questions!
ChatGPT isn’t static—it works by calculating probabilities based on your input. This means it’s not always going to give the same answer every time.
The randomness depends on a setting called ‘temperature,’ which controls how creative the responses are. A higher temperature means more variety, but it can also lead to inconsistent answers.
It’s because ChatGPT is trying to simulate intelligence rather than follow a fixed script. This leads to variations in its answers.
On top of that, there’s something called ‘hallucination,’ which is when it confidently gives wrong information. Developers are working on fixing this, but it’s still a known issue.
Have you tried using OpenAI’s playground? There, you can adjust settings like temperature. If you lower it, the answers become more consistent.
You could also experiment with something like maximum lexical density to make the responses sharper.
There’s a big difference between answers that are different and answers that are wrong. Which one are you dealing with here?
Haha, maybe if you had one more reviewer—1201 instead of 1200—GPT would magically become perfect. Just kidding!