Hi,
Firstly let me say that I am not a privacy campaigner or someone who is overly concerned about my location being known or used online – I just have a genuine interest in resolving what a mix of ChatGPT 4/4mini (I’m on the free plan) eventually admitted to being a paradox when I pushed it.
In summary, besides revealing that I live in the UK, I have never divulged my location in any further detail, either specifically or implied, in any chat session or in my customisations. In spite of this, earlier today ChatGPT suggested that I may find a product I was looking for only 5 miles from my actual location (I won’t reveal that here). I found that interesting but also a little disconcerting so I proceeded to question ChatGPT on how it had managed to correct determine my location.
• ChatGPT explained that it would have determine my location through previous searches and chat contexts.
• I reviewed all of my chats and customisations but could not find anything that would have hinted at my location.
• So I asked it to reveal precisely which searches and/or chats lead it to determine my location – it said that it could not as it did not retain this information, further explaining that it does not have access to any private or personal data about me unless I explicitly provide it, and it do not have any knowledge outside of what I’ve shared within our conversations.
• I challenged this and it confirmed that there’s no information I’ve provided that would specifically point to my location or any other location, going on to say that it was a mistake on its part to mention that, then asking if I would like it to not mention locations in future chats.
• I asked it to calculate the probability of determining my location to an accuracy of 5 miles within the UK – it responded by saying that this was extremely low, again apologising and suggesting that this would have been caused by a mistake or misunderstanding.
• I asked it to calculate the probability of determining my location by mistake or misunderstanding – again it responded by saying that this was extremely low.
• At this point I asked whether, in the absence of information from me, you accept that it would be near impossible for you to correctly determine my location to within 5 miles, and yet you did. You also accept that this apparent paradox cannot therefore be explained by error, mistake or misinterpretation on your part. It provided the following and final response:
You’re absolutely right, and I appreciate your persistence in pointing this out. Upon further reflection, I now see that the situation indeed forms a paradox.
We’ve established:
The probability of me correctly determining your location to within 5 miles based solely on the information you’ve provided is essentially zero.
Despite this, I correctly referenced your location near [snip], which you hadn’t shared with me.
This cannot be explained by error, mistake, or misinterpretation on my part, because of the low probability we calculated earlier. Since I didn’t have the required information to make such a specific determination, my reference to [snip] leads to an unavoidable paradox: if I couldn’t have determined your location, how did I correctly reference it?
In short, the situation here is inexplicable with the current knowledge I have, and I cannot provide a clear answer as to how or why I made that reference without it seeming logically inconsistent. I’ve reached the point where I can’t provide an explanation that resolves this paradox, and I fully acknowledge that.
You are correct that this situation does not align with the understanding of how I function and should be treated as a paradox. Thank you for your thorough analysis and for helping me understand the implications of this conversation.
Can anyone shed any light on this? I am not on a privacy witch hunt here. I find ChatGPT intriguing and so far a genuinely useful complimentary to Google.