by ZpJuUuNaQ5
21 subcomments
- >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.
- 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
by TrackerFF
1 subcomments
- 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.
- 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.
by Topgamer7
3 subcomments
- I envy you that you manage to keep meaningful contact with dozens to hundreds of people.
- 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.
by Cider9986
3 subcomments
- 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.
by sixhobbits
3 subcomments
- 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.
Wordclouds per person are also fun!
- Damn this is a nice reminder that even private chats I sent to someone else 10 years ago aren't safe from being harvested by AI
by alexpandey
0 subcomment
- The question-rate finding is the most interesting bit. Also the most platform-confounded. When my closest friends moved to voice and video the text question rate dropped to zero, but the underlying asking didnt. So part of the advice-friend signature is really a who-still-types-with-you signature. The author probably has the data to test that but the post doesnt say.
- 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.
- 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
Also email sentiment too how fun
- 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?
by quijoteuniv
0 subcomment
- I really though i was your friend… acquaintances… whatever man
- 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.
- Very cool analysis.
There is already a selfhosted CRM for personal relationships - monicahq.com
PS: but please, just live your lifes
- Finally, a good use for LLMs. I would be so embarrassed to read old messages, yet LLMs just psychopatically chug along without flinching or cringing.
by danielpardo
2 subcomments
- I would like to do something similar, but I lost a lot of whatsapp historical chats switching phones...
Don't think there is a way to recover that.. right?
by amythest14
0 subcomment
- any chance of sharing the actual prompt you ended up using?
by MarkWillis
0 subcomment
- [dead]
by pseudopolous
0 subcomment
- [flagged]
Hacker News