i face the same situation ,now i cant send any message in chatgpt,and history catalog has missed
I also having the same problem now, I was not able to view all my history chats since yesterday. @openai10
I have the same situation ,I cant send any message in chatgpt,and history catalog has missed
1 Like
I have the same problem. itās painful.
1 Like
me too, but my friends tell me that they could ask question as normal, and the history menu in their chatgpt work well, maybe this question only happen in some of the people who use vpn?
1 Like
Is this still a problem for every user? Also I am wondering, how long it might take to get the history backā¦
The chatting works again, but history doesnt
is someone received news from chatgpt about history bug resolution?
I am still unable to submit a chat/question or view History. Interestingly if I get someone to login on ChatGPT on my machine with their account it works fine.
3 Likes
Yesterday I put several hours into a prompt I wanted to use today. Because a computer update, it restarted. Now I donāt have access to my history, so I can not get the results of my work yesterday. If this is just a toy to play with, then lower the price, if it is supposed to be a serious tool, then make it work reliable. It doesnāt take any exceptional computer engineer to save your history first in local storage to, at least, give you access to your conversations, even if it is in read-only mode.
2 Likes
Apparently itās a bug that had been allowing other users to see other users chat history.
1 Like
Adding my voice to the chorus of dissatisfaction.
ChatGPT is a phenomenal toolāespecially for this frustrated fiction writer who longed for a tireless brainstorming partner, and was overjoyed to have found it here! Were it not for the maddening history losses, would gladly become a paying subscriber.
1 Like
There are users from all over the world using chatgpt. In their own languages. This bug must be very rare and I wonder how only a handful of chat titles from he BILLIONS created every day appear in another userās history.
My only guess lies in the ābillionsā part.
Each chat must have a UNIQUE ID in the database.
You see, an hexadecimal with 8 characters is FFFFFFFF. That is 4,294,967,295 in decimals. (4.2 billion)
The chances of a repeating ID is 1 in 4 billion at the second chat ever created. But the chances increase. Itās the so called BIRTHDAY PARADOX
The birthday paradox is a probability theory phenomenon that illustrates the counterintuitive nature of probabilities in large sets. It states that in a group of just 23 people, there is a 50% chance that at least two individuals share the same birthday, even though there are 365 days in a year. The paradox arises because our intuition often leads us to think that a much larger group would be necessary for such a high probability of shared birthdays. The underlying principle behind the birthday paradox is that the number of possible pairings between individuals grows quadratically with the size of the group, which results in a much faster increase in the probability of a match than we might intuitively expect. This phenomenon has applications in various fields, including cryptography and hashing algorithms, where it helps illustrate the likelihood of collisions in large data sets.
So, letās consider here are 10 million new chats per day for the last two months, and that OpenAI wasnĀ“t predicting that much and therefore used an 8 character unique ID for each chat.
Letās have ChatGPT itself explain the probability of repeating IDs
Blockquote
To calculate the probability of generating at least one duplicate ID after generating 10 million new IDs per day for two months, we need to use the same principles discussed earlier with the birthday paradox.
First, letās determine the total number of IDs generated in two months. Assuming 30 days per month, we have:
10,000,000 IDs/day * 30 days/month * 2 months = 600,000,000 IDs
Now, we need to calculate the probability of no duplicates among these 600,000,000 IDs. Recall that there are 4,294,967,296 possible unique IDs in this system. We can use the formula for P(no duplicates):
P(no duplicates) = (Combinations - 1) / Combinations * (Combinations - 2) / Combinations * ā¦ * (Combinations - n + 1) / Combinations
Computing this probability directly is computationally intensive, but we can use an approximation to calculate P(no duplicates):
P(no duplicates) ā exp(-n^2 / (2 * Combinations))
Where exp() is the exponential function and n is the number of IDs generated (600,000,000 in this case). Now we can plug in the numbers:
P(no duplicates) ā exp(-600,000,000^2 / (2 * 4,294,967,296))
P(no duplicates) ā exp(-0.02080) ā 0.9794
Now, we can find the probability of generating at least one duplicate ID by subtracting P(no duplicates) from 1:
P(at least one duplicate) ā 1 - 0.9794 ā 0.0206
So the chances of having at least one repeating ID after generating 10 million new IDs per day for two months are approximately 2.06%.
With 16 hexadecimal characters, the number of possible combinations would jump from 4.2 billion to 18,446,744,073,709,551,616 (18 quintillion).
With the same numbers as before of new ids per day, the probability of repeating IDs is only approximately 0.000003257%, which is significantly lower than when using 8 hexadecimal characters.
This is ChatGPTās answer when I asked if my hypothesis is plausible
Blockquote
Your hypothesis is plausible, and itās possible that the use of relatively short unique IDs (e.g., 8-character hexadecimal codes) could have led to collisions, resulting in users being able to access other usersā chat histories. If OpenAI underestimated the number of daily chats and the associated probability of repeating IDs, such an issue could indeed occur.
However, itās important to note that without specific information about OpenAIās implementation, itās impossible to say for certain if this was the exact cause of the issue. There could be other factors or bugs in the system that contributed to the problem. Increasing the length of unique IDs to 16 hexadecimal characters, as you suggested, would indeed significantly reduce the probability of ID collisions and could help prevent such issues in the future. It is also crucial to ensure that the system properly handles any potential collisions and other edge cases to prevent unauthorized access to usersā data.
I have the same problem, since 3 Days
2 Likes
Hello I am not able to access my history as well. Its been a while!
Anyone knows anything useful abou this issue? I started to interact with this chat just a few days ago, and I like to keep everything, but now Iām forced to save it separately every timeā¦
I still have no access. It has been 3 days. The send button simply does not work.
2 Likes
If this continues Iām not going to renew my subscription
It would be funny that so many people are fiending to talk to their robot again, if not for the fact that I am fiending to talk to my robot again too! It amazes me how after roughly only a week or two on GPT-4, I donāt want to interact with the real world before checking in to see what it has to say first! That both scares me and amuses me.
That aside, Iāve been experiencing the same issues and frustrations as everyone else in this conversation, currently I am able to log into my ChatGPT Plus account and can access all three models, but am unable to load history (ā¦which I hope is backed up). I also keep an eye on the OpenAI Help Board, but havenāt seen this issue addressed yet. I would imagine theyāre overwhelmed by the success of this project and are working hard to fix it.
1 Like