thoughts on LLMs for Social Networks
💠in the future, your LLM will talk to my LLM, and neither of us will benefit from the sage information exchanged.
I find myself leaning towards social networks are bad, and I shouldn't work on things opposed to the concept of having less of them these days.
people like engagement. but, there is the iron law of calculus here: the amount of time people spend with your content is equal to the time people spend with your content.
if you want the endorphins of 15 people liked this post, you need 15 people to read and like the post.
🔥 but, if the machine can score it ...
💡 "likes" mean many things. i went to the baseball game. LIKE.
🔥 i have seen no evidence that a social network with xantham functionality would be useful to anyone.
we live in an age where censorship of social networks, while correct, is also unpopular enough to kill an app.
the LLM-powered social network will not be a social network. not the way you know it.
you talk to it. it remembers. and, nicely, keeps some public information about you available.
the panopticon of you know who knows what will collapse. 💡( at least for us, this is a good thing.)
the machine can have all your photos, and know when to broadcast one. 🔥( it would be nice to offload the responsibility of knowing whether to share that the pans are cleaner)
if Zuck calls me, I can read him this email. 💡( it would have to be Zuck. all the other social networks are run by committee, or by ketamine-trip)
I don't see how LLMs can handle given person A (with dossier) and person B (with dossier) and COMM_HISTORY, send a message about the past month.
that feature is the lynchpin for any tool to replace Facebook.
🔥 it's my newslettr. again.
💠the solution is never the newslettr.