I wanted to do this so bad back in like 2009. I archived MSN messenger chats and my texts and everything and I figured sometime in the future I'd be able to analyze them and maybe even train a chat bot to imitate me. But over the years I lost hard drives and cell phones and accidentally deleted stuff and yeah it didn't happen. Very cool this guy was able to do it
I mostly text on Signal with disappearing messages so I wouldn't be able to do this. Most people are fine with disappearing messages at 4 weeks, but a few people like to keep their chats forever.
I do wanna do the same, but at times I fear about getting too aware from the insights of the analysis, I fear my opinions my faith might change in certain people.
I do some similar charting etc with telegram data dumps that you can still get from the "telegram lite" desktop app even though they removed the export functionality from the main app.
For removing noise you might want to look into TF-IDF instead of the manual method described in the post that I didn't understand. It basically looks for words common across the whole corpus as noise or ones that appear within a specific chat much higher than the whole dataset as interesting.
You can also do some fun stuff by finding phrases used asymmetrically eg more by one person in the convo than the other, or over time.
I don't understand how some people write hundreds of text/chat messages per day. I am communicating by talking to people almost 100% of non-work, most of the discussions are face to face, I write or receive a handful of texts or chats per week, maybe a dozen per month.
I find text messages impersonal and it also takes longer to communicate clearly what we need. There is so much lost. Even chats and emails for work are at risk of creating misunderstanding, especially because English is not the native language of most of my coworkers, all these adds to result in pretty low quality communication.
>15 close friends, 50 regular contacts, 150 active acquaintances
Seems like a lot. For me, it's 0 close friends, 0 regular contacts, and 0 active acquaintances. I think I simply never developed any useful social skills which would help me make and keep friends or acquaintances. I wish I had (somehow) kept in touch with at least some people I have met throughout my life. It has never been easier to stay in contact than in all of human history, but no, I had to ghost and ignore everyone and everything. After 29 revolutions around the sun, I have only now started to realize that all that vacuous superiority has led me nowhere. There is only a faint aftertaste of missing out, which sticks to me like tar. I can’t wash it off.
Dude, you can still recover some of those friendships. I went through a phase during college where I alienated a TON of my friends, and that was on me. For a variety of reasons we don't need to expound upon here. I eventually reached out to them, made my apologies, and would occasionally check in and see how they were doing. A few of those friendships are repaired and growing again. I think that there's a tendency toward second chance for people who've recognized their shortcomings, worked on them, and then taken responsibility for their actions.
As a hypothetical example, what happens if you could train a custom LLM with your own chats. And it responds to people, and gives the same response as you would. Would that be a form of mental cloning?
The "Sasha" section brought back a load of memories from my childhood. As an Alex growing up in Western Europe with no connections to anything East it was just my Russophile father that used to call me Sandy or Sasha some of the time.
One byproduct of me storing all my harddrives going back to 2001, is that I also have all my chat logs. Last fall I decided to invest in some HD adapter, as I was looking for pictures of my friend for his 40th birthday - miraculously all drives worked, though some would just power off at random intervals.
I decided to look through old IRC chats, and I honestly couldn't for sure say who maybe 30% of the people are just based on their nicknames. Nicknames I hadn't seen in 20 years. So I decided to feed all the raw chats into a LLM, and see if the model could string together names and nicknames. It managed to do surprisingly well! Many of the chats were not personal chats, but from the channels. My most active channels were local groups from my small town, so the names would naturally be named there, but I couldn't be bothered with sifting through the vast amounts of chats/text.
I then noticed that in the mid 2000s, MSN Messenger really took off, so most my chats were done there. Or ICQ if I were chatting with people from the US.
Then, around 2009/2010 Facebook became the standard (though it seems like my account was created in 2007), and most chats are via messenger.
This is a great idea, I was leaning towards doing it with openrouter since the cost now is so affordable then codex but seeing someone else do it, can do it too with WhatsApp, iMessages, facebook, scattered irc logs
Why do people keep chats? I have none, oldest is 6 months which is when I set my chat app to delete them. I'm the same with emails. Genuinely don't understand why you'd want to keep this stuff. I'd never go back and reread them.
Might depend on how old you are. If you're in your 20s or 30s, you probably still remember the silliest details like a childhood friend's middle name. But after that, bright memories start to fade, and trust me, it only gets worse. That's when you realize that having a little bit of memory scaffolding like old email, chats, and calendar is helpful for once-easy questions like "where did we go camping that one time when I was 28 or 29?"
You will wish you had them when you need them. There might be reasons in the future that you don't realize now. Plus the fact that storage is cheap. So the most logical thing to do is to keep them unless you have a good reason to delete.
28 comments
[ 3.1 ms ] story [ 54.0 ms ] threadDon't think there is a way to recover that.. right?
For removing noise you might want to look into TF-IDF instead of the manual method described in the post that I didn't understand. It basically looks for words common across the whole corpus as noise or ones that appear within a specific chat much higher than the whole dataset as interesting.
You can also do some fun stuff by finding phrases used asymmetrically eg more by one person in the convo than the other, or over time.
Wordclouds per person are also fun!
I find text messages impersonal and it also takes longer to communicate clearly what we need. There is so much lost. Even chats and emails for work are at risk of creating misunderstanding, especially because English is not the native language of most of my coworkers, all these adds to result in pretty low quality communication.
Seems like a lot. For me, it's 0 close friends, 0 regular contacts, and 0 active acquaintances. I think I simply never developed any useful social skills which would help me make and keep friends or acquaintances. I wish I had (somehow) kept in touch with at least some people I have met throughout my life. It has never been easier to stay in contact than in all of human history, but no, I had to ghost and ignore everyone and everything. After 29 revolutions around the sun, I have only now started to realize that all that vacuous superiority has led me nowhere. There is only a faint aftertaste of missing out, which sticks to me like tar. I can’t wash it off.
I decided to look through old IRC chats, and I honestly couldn't for sure say who maybe 30% of the people are just based on their nicknames. Nicknames I hadn't seen in 20 years. So I decided to feed all the raw chats into a LLM, and see if the model could string together names and nicknames. It managed to do surprisingly well! Many of the chats were not personal chats, but from the channels. My most active channels were local groups from my small town, so the names would naturally be named there, but I couldn't be bothered with sifting through the vast amounts of chats/text.
I then noticed that in the mid 2000s, MSN Messenger really took off, so most my chats were done there. Or ICQ if I were chatting with people from the US.
Then, around 2009/2010 Facebook became the standard (though it seems like my account was created in 2007), and most chats are via messenger.
Also email sentiment too how fun
PS: but please, just live your lifes