# How to Survive AI’s Social Media Revolution in 2025

Source: https://www.youtube.com/watch?v=21ycFMBq5js
Recap page: https://rapidrecap.app/video/21ycFMBq5js
Generated: 2025-08-21T13:32:54.591+00:00

---
## Quick Overview

AI-generated avatars can now be created from single photos and audio, enabling infinite-length videos that sync speech with facial expressions, potentially revolutionizing content creation and communication.

**Key Points:**
- StableAvatar technology enables the creation of infinite-length audio-driven avatar videos from single photos, syncing speech with facial expressions.
- AI models are being developed to understand and translate concepts across different data types and languages, mapping relationships between them.
- The concept of 'doompompting' is introduced as a shift towards more active AI interaction, potentially leading to a sense of felt productivity and intellectual partnership.
- Research is questioning the true nature of AI reasoning, particularly "chain-of-thought" reasoning, and whether it truly mimics human cognitive processes.
- AI advancements are impacting various industries, including content creation (avatar videos) and potentially Hollywood, raising questions about future job markets and societal changes.
- The discussion touches upon the ethical considerations of AI, including its role in surveillance and the potential for AI to shape human behavior and thought patterns.
- The video highlights the growing trend of "wealth migration," where millionaires are moving to countries offering more favorable economic and lifestyle conditions, impacting global economies.

![Screenshot at 00:03: AI-generated avatars are showcased, demonstrating their ability to sync speech with facial expressions.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-00-03.png)

**Context:** The video explores recent advancements and discussions in the field of Artificial Intelligence, focusing on AI-generated content, reasoning capabilities, and the societal impact of AI. It references specific research papers and projects, as well as popular content creators and their work, to illustrate the rapid progress and evolving landscape of AI technology.

## Detailed Analysis

This video introduces StableAvatar, a new AI model capable of generating infinite-length audio-driven avatar videos from a single photo. The process involves using Midjourney for image generation, Google Veo 3 for animation, and Suno for music, with the entire pipeline assembled in DaVinci Resolve. The technology allows solo creators, small studios, and anyone with a story to create films using AI.  The paper "Harnessing the Universal Geometry of Embeddings" by Rishi Jha et al. is discussed, highlighting how AI models can learn to translate embeddings without paired data and represent concepts in a universal latent space. The model learns to figure out the distance between two concepts and how they are exposed to different data. The video also touches upon the concept of "doompompting" as the new "doomscrolling," where AI interaction shifts from passive consumption to active creation, and the dopamine of likes and follows becomes the dopamine of felt productivity, social validation, and intellectual partnership.  The presenter also references a YouTube channel called "Wes and Dylan" and their podcast, where they interviewed Matt Wolfe about superintelligence and AI's impact on Hollywood. They also mention meeting other AI creators and discussions around AI ethics and potential future scenarios involving AI. The demonstration of StableAvatar's capabilities includes generating videos from single images, showcasing realistic facial expressions, and creating multi-character portrait animations with expression-augmented diffusion transformers. The paper "Is chain-of-thought AI reasoning a mirage?" by Sean Goedecke is also referenced, questioning the true nature of AI reasoning.

### StableAvatar AI

- Generates infinite-length audio-driven avatar videos from single photos
- Uses Midjourney, Google Veo 3, Suno, and DaVinci Resolve
- Enables creators to make AI-driven films

### AI Telepathy Research

- Paper "Harnessing the Universal Geometry of Embeddings" explores AI's ability to understand relationships between concepts
- Models learn from unpaired data and translate text embeddings
- Features vector space translation and latent representations

### Doompompting vs. Doomscrolling

- Discusses the shift from passive consumption to active creation through AI
- AI interaction leads to dopamine from productivity, validation, and intellectual partnership

### AI Ethics and Future

- Mentions podcasts discussing AI's impact on society, superintelligence, and ethical considerations
- Explores AI's potential to disrupt industries like Hollywood and surveillance

### FantasyPortrait AI

- Enhances multi-character portrait animation with expression-augmented diffusion transformers
- Creates realistic facial expressions and transitions
- Demonstrates AI's capability to generate diverse and expressive avatars

![Screenshot at 00:03: AI-generated avatars are showcased, demonstrating their ability to sync speech with facial expressions.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-00-03.png)
![Screenshot at 00:32: A trailer for 'Beast Games: 2082' is shown, highlighting AI-generated visual content.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-00-32.png)
![Screenshot at 01:15: The YouTube channel 'Wes and Dylan' is displayed, featuring content related to AI and superintelligence.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-01-15.png)
![Screenshot at 01:31: A video demonstrating "AI Telepathy" explores vector embeddings and latent representations.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-01-31.png)
![Screenshot at 01:38: An article titled "Doomprompting Is the New Doomscrolling" discusses AI's impact on human behavior.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-01-38.png)
![Screenshot at 01:43: A research paper titled "StableAvatar: Infinite-Length Audio-Driven Avatar Video Generation" is presented.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-01-43.png)
![Screenshot at 01:53: Examples of audio-driven avatar video generation are shown, showcasing various faces and expressions.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-01-53.png)
![Screenshot at 02:01: A paper questioning AI reasoning, "Is chain-of-thought AI reasoning a mirage?", is highlighted.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-02-01.png)
![Screenshot at 02:09: A progress chart for AGI and ASI \(Artificial General Intelligence and Artificial Superintelligence\) is displayed.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-02-09.png)
![Screenshot at 03:15: A group photo with AI creators, including those from the 'Wes and Dylan' podcast, is shown.](https://ss.rapidrecap.app/screens/21ycFMBq5js/00-03-15.png)