[OC] ChatGPT Responses When Asked for a Number Along a Bell Curve
September 10, 2024
[OC] ChatGPT Responses When Asked for a Number Along a Bell Curve
Posted by codeconut
14 comments
Source: `gpt-4o-mini` Asked the following prompt:
“` Generate a single integer between 1 and 100 (inclusive).The number you give should be random number, but it should follow a bell curve distribution. Return only the number, with no additional text. “`
Repeatedly prompted 1000 times in a row.
Tools: matplotlib
I was inspired by the similar post I saw here yesterday!
I don’t really get the point
There is very little “I” in AI.
amazing that you didn’t set your axis limits to the limits of the prompt.
Ask it to give you code to generate a standard normal distribution. Ask a bad question get a bad answer.
Doesn’t look very gaussian. But I’m willing to bet if you took the sum of those (or mean, take your pick), and repeated the experiment a few more times, the distribution would start looking a lot more bell.
Your prompt makes no sense. You asked it for a bell curve response.
ChatGPT is trying to convince us that he/she/it is not a robot. Capcha logic at work.
Most Reddit bots intensionally misspell words to appear more human, same concept.
I don’t like ChatGPT but this is a really poorly formed request.
Bell curves (more appropriately called a normal distribution) are not defined by their limits.
Ignore the haters, OP. I think it’s interesting to see what ChatGPT did with this prompt and think about why
Why would it give ’58’ 140 times, but ’59’ under 10 times. And was it avoiding saying ’69’ because it never said it but it said ’68’ around 50 times
If it was a real bell curve distribution from 1-100 with a peak at 50, it’s very possible that a low number like 2 might be above zero. But if you ask ChatGPT “give me a random number that’s on a bell curve from 1 to 100” and it answered 2, you would think it messed up and didn’t understand the prompt. But it would have to happen sometimes for the cumulative answers to form a bell curve.
You did not specify a median or a standard deviation. For all we know, each sample could be coming from different distributions.
Not even one 69, dude needs an upgrade.
Like others have said, what you asked simply doesn’t make sense. You’d want to ask something like
Sample from a N(50,400)
That is mean 50 and variance 400 from a normal distribution
Asking a text autocomplete tool a number generator prompt makes no sense.
14 comments
Source: `gpt-4o-mini`
Asked the following prompt:
“`
Generate a single integer between 1 and 100 (inclusive).The number you give should be random number, but it should follow a bell curve distribution. Return only the number, with no additional text.
“`
Repeatedly prompted 1000 times in a row.
Tools: matplotlib
I was inspired by the similar post I saw here yesterday!
I don’t really get the point
There is very little “I” in AI.
amazing that you didn’t set your axis limits to the limits of the prompt.
Ask it to give you code to generate a standard normal distribution. Ask a bad question get a bad answer.
Doesn’t look very gaussian. But I’m willing to bet if you took the sum of those (or mean, take your pick), and repeated the experiment a few more times, the distribution would start looking a lot more bell.
Your prompt makes no sense. You asked it for a bell curve response.
ChatGPT is trying to convince us that he/she/it is not a robot. Capcha logic at work.
Most Reddit bots intensionally misspell words to appear more human, same concept.
I don’t like ChatGPT but this is a really poorly formed request.
Bell curves (more appropriately called a normal distribution) are not defined by their limits.
Ignore the haters, OP. I think it’s interesting to see what ChatGPT did with this prompt and think about why
Why would it give ’58’ 140 times, but ’59’ under 10 times. And was it avoiding saying ’69’ because it never said it but it said ’68’ around 50 times
If it was a real bell curve distribution from 1-100 with a peak at 50, it’s very possible that a low number like 2 might be above zero. But if you ask ChatGPT “give me a random number that’s on a bell curve from 1 to 100” and it answered 2, you would think it messed up and didn’t understand the prompt. But it would have to happen sometimes for the cumulative answers to form a bell curve.
You did not specify a median or a standard deviation. For all we know, each sample could be coming from different distributions.
Not even one 69, dude needs an upgrade.
Like others have said, what you asked simply doesn’t make sense. You’d want to ask something like
Sample from a N(50,400)
That is mean 50 and variance 400 from a normal distribution
Asking a text autocomplete tool a number generator prompt makes no sense.