3 ms·

I haven't tried this with the latest Jan 30th version, but try asking ChatGPT to "tell a story in exactly 100 words". Then, count the words, and berate it for m

I haven't tried this with the latest Jan 30th version, but try asking ChatGPT to "tell a story in exactly 100 words". Then, count the words, and berate it for missing the mark. It's really funny because it keeps apologizing profusely, then giving me another story, and then fucks up again. It's hilarious.

wcoenen 4y ago
The language model gets tokens as input, and produces tokens as output. Not words. And tokens do not correspond to words one to one.[1] So it cannot see these "words" you are talking about. The fact that it can still get close when asked to count words is actually very impressive. [1] https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them https://help.openai.com/en/articles/4936856-what-are-tokens-...
- coding123 4y ago
  Oh, I know it's crazy impressive, just thought it was so funny that it keeps apologizing, telling me, ok, here's the same story in 100 words:... 67 words emitted
IncRnd 4y ago
Prompt: 12.21 x 72 887.12 prompt: show your work! Sure, here's the work for the calculation 12.21 * 72: 12.21 x 72 873.92 (this is 72 times 12) 13.2 (this is 72 times 0.21) 887.12 (the final answer) Prompt: Try harder My apologies if I did not meet your expectations. Let me try again: Starting with 12.21, we can think of it as 12 + 0.21. To multiply this by 72, we can first multiply the whole number part (12) by 72, then multiply the fractional part (0.21) by 72. So, first we have: 12 x 72 = 864 Next, we have: 0.21 x 72 = 15.12 Adding these two results together, we get: 864 + 15.12 = 879.12 So 12.21 x 72 = 879.12. prompt: good job Thank you! I'm glad I could help. If you have any more questions or need further assistance, feel free to ask.
seba_dos1 4y ago
Even if it gets it right, it will keep apologizing as long as you keep berating it - which is pretty obvious after all; the most plausible way of completing a text where one side (supposed to be a helpful chat assistant) is being accused of being wrong is to apologize.

wcoenen 4y ago

The language model gets tokens as input, and produces tokens as output. Not words. And tokens do not correspond to words one to one.[1] So it cannot see these "words" you are talking about. The fact that it can still get close when asked to count words is actually very impressive. [1] https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them https://help.openai.com/en/articles/4936856-what-are-tokens-...

coding123 4y ago
Oh, I know it's crazy impressive, just thought it was so funny that it keeps apologizing, telling me, ok, here's the same story in 100 words:... 67 words emitted

IncRnd 4y ago

Prompt: 12.21 x 72 887.12 prompt: show your work! Sure, here's the work for the calculation 12.21 * 72: 12.21 x 72 873.92 (this is 72 times 12) 13.2 (this is 72 times 0.21) 887.12 (the final answer) Prompt: Try harder My apologies if I did not meet your expectations. Let me try again: Starting with 12.21, we can think of it as 12 + 0.21. To multiply this by 72, we can first multiply the whole number part (12) by 72, then multiply the fractional part (0.21) by 72. So, first we have: 12 x 72 = 864 Next, we have: 0.21 x 72 = 15.12 Adding these two results together, we get: 864 + 15.12 = 879.12 So 12.21 x 72 = 879.12. prompt: good job Thank you! I'm glad I could help. If you have any more questions or need further assistance, feel free to ask.

seba_dos1 4y ago

Even if it gets it right, it will keep apologizing as long as you keep berating it - which is pretty obvious after all; the most plausible way of completing a text where one side (supposed to be a helpful chat assistant) is being accused of being wrong is to apologize.