3 ms·

Oh - I have only been to Hong Kong. Actually I manually added China to the list and left the total 30-35% alone which stat I got from the perplexity.ai (I assum

Oh - I have only been to Hong Kong. Actually I manually added China to the list and left the total 30-35% alone which stat I got from the perplexity.ai (I assumed it was a hallucination). I have edited the comment back tot he original. Thanks for the clarification.

grues-dinner 2y ago
One does wonder, if an AI is so untrustworthy that you pre-emptively hypercorrect(1) its output because you assume it's hallucinating, what the point is of using it to write an authoritative-sounding comment in the first place. (1): https://en.wikipedia.org/wiki/Hypercorrection https://en.wikipedia.org/wiki/Hypercorrection
- librasteve 2y ago
  Artificial Intelligence meets Human Stupidity (frankly I am too lazy to manually fact check the definitive list of RHD countries for the purpose of a simple HN post - but I do want to get across that 30-35% of humanity is not just the UK - so I reached for the LLM and ended up thinking (hmmm - I have been to HK and they were driving on the left) and (hmmm - looks like the AI forgot China which it probably doesn't realise is a "notable" country) - so I put China in the list, was furthermore too lazy to redo the math so stuck with 30-35% since that was the "safe" choice). I submit that the real damage was done by using an "authoritative-sounding" set of words to present my seat of the pants guestimate and for this I stand corrected.
rsynnott 2y ago
So, I'm fascinated, what's the motive for leaving these comments which are just a (frequently questionably correct) summary from an LLM? Like, if people want to ask a stochastic parrot, they can presumably ask one themselves; commenting with the output of one seems like a weird behaviour.
- librasteve 2y ago
  I felt that the prior comments were very focused on Britain (and, as a counterpoint, our European neighbours who drive on the right) and that there is a global aspect to the topic which would be appreciated by the global HN audience. Therefore I thought it would be beneficial to the discussion to bring in that a largish fraction of the world population lives in countries that drive on the left. Since I didn't know what that fraction is and I myself was curious to learn it, I used a popular AI search utility to find the information. When I had gathered the information I thought it was relevant to the topic and informative and that the best way to share it was to copy it into the post text field. I'm not sure that my answer is that fascinating, but I believe that my motive was well intentioned and that the method was not unreasonable. In hindsight I should have cited my source so that others did not take an "authoritative-sounding" piece of text as gospel and that my role had been selection and a cursory edit. Funnily enough I have been quite resistant to LLMs, but I am finding that they are a much better way to do research on the web than Google. I also believe that the reason many folks think that LLM is a step towards AGI is that humans are also stochastic parrots. Whereas parrots are not.

grues-dinner 2y ago

One does wonder, if an AI is so untrustworthy that you pre-emptively hypercorrect(1) its output because you assume it's hallucinating, what the point is of using it to write an authoritative-sounding comment in the first place. (1): https://en.wikipedia.org/wiki/Hypercorrection https://en.wikipedia.org/wiki/Hypercorrection

librasteve 2y ago
Artificial Intelligence meets Human Stupidity (frankly I am too lazy to manually fact check the definitive list of RHD countries for the purpose of a simple HN post - but I do want to get across that 30-35% of humanity is not just the UK - so I reached for the LLM and ended up thinking (hmmm - I have been to HK and they were driving on the left) and (hmmm - looks like the AI forgot China which it probably doesn't realise is a "notable" country) - so I put China in the list, was furthermore too lazy to redo the math so stuck with 30-35% since that was the "safe" choice). I submit that the real damage was done by using an "authoritative-sounding" set of words to present my seat of the pants guestimate and for this I stand corrected.

rsynnott 2y ago

So, I'm fascinated, what's the motive for leaving these comments which are just a (frequently questionably correct) summary from an LLM? Like, if people want to ask a stochastic parrot, they can presumably ask one themselves; commenting with the output of one seems like a weird behaviour.

librasteve 2y ago
I felt that the prior comments were very focused on Britain (and, as a counterpoint, our European neighbours who drive on the right) and that there is a global aspect to the topic which would be appreciated by the global HN audience. Therefore I thought it would be beneficial to the discussion to bring in that a largish fraction of the world population lives in countries that drive on the left. Since I didn't know what that fraction is and I myself was curious to learn it, I used a popular AI search utility to find the information. When I had gathered the information I thought it was relevant to the topic and informative and that the best way to share it was to copy it into the post text field. I'm not sure that my answer is that fascinating, but I believe that my motive was well intentioned and that the method was not unreasonable. In hindsight I should have cited my source so that others did not take an "authoritative-sounding" piece of text as gospel and that my role had been selection and a cursory edit. Funnily enough I have been quite resistant to LLMs, but I am finding that they are a much better way to do research on the web than Google. I also believe that the reason many folks think that LLM is a step towards AGI is that humans are also stochastic parrots. Whereas parrots are not.