메타 특허 인터넷 봇 사후에도ㅡ 이수만

[마운틴듀2]

https://www.facebook.com/share/1QtaWbrnwV/

Facebook

Meta patented AI that can learn your online behavior, simulate your digital persona, and keep “you” interacting on social media even after you die‼️...👀👇

This patent, US12513102B2, is a Meta Platforms patent titled “Simulation of a user of a social networking system using a language model.” In simple terms, it describes creating an AI model that learns how a particular person behaves on a social network, then allowing an automated bot to interact with other people as if it were that person. Meta is listed as the assignee, and the patent was granted on December 30, 2025.

The basic idea is to take a person's past activity—such as posts, comments, likes, shares, messages, interests, relationships, and other interaction patterns—and use that information to customize or retrain a language model. The resulting model is intended to predict how that specific person would respond to new content. A bot could monitor the person's newsfeed, send relevant content to the language model, receive a predicted response, and then perform that interaction on the user's behalf.

For example, if someone normally comments on certain subjects in a particular writing style, the system could learn those patterns. When a new post appears, the Al could be prompted with that post and asked what the person would probably do-perhaps comment, like it, share it, or respond in some other way. The patent specifically says the model is trained from the target user's own previous interactions so its responses resemble what that person would have produced.

One of the most striking parts of the patent is that this is not limited to helping someone while they are busy or temporarily offline. It explicitly states that the technology could simulate someone who has been absent for a long time and could continue simulating a person after that person has died. The patent says this could allow other users to continue experiencing the person's presence on the social network even though the person is deceased.

The patent also goes beyond simply analyzing someone's written posts. It describes user profiles containing information such as work history, education, interests, hobbies, preferences, location, photos, videos, relationships, and social connections. Activity logs can contain information about comments, likes, shares, events, groups, communication with other users, and other behaviors. It even discusses information supplied by consenting third-party systems, including things such as external website activity, purchases, advertisements interacted with, webpage viewing patterns, email, messages, and voice communications, which could potentially contribute to the model's training data.

So, in plain language, the patent describes building an Al-based digital behavioral copy of a social-media user. It would learn from that person's accumulated digital history and use a language model to predict how the person would react, communicate, or interact online. Once trained, an automated bot could effectively maintain that person's social-media activity when they were unavailable-and, according to the patent itself, potentially continue representing that person's online personality after death.

Simulation of a user of a social networking system using a language model
https://patents.google.com/patent/US12513102B2/en