Why Midjourney was created? And The Man Behind it

The paragraph begins with a personal query to the founder of Midjourney about the reasons for creating the platform, its long-term vision, and why its images are exceptional. The author highlights the lack of information, marketing, and interviews with the founder, David Holtz, which sparked curiosity. Before revealing what they discovered, the author sets the context by explaining why Midjourney is significant, with its massive Discord user base and GPU usage. It emphasizes that David Holtz's reputation in the tech field allowed Midjourney to scale rapidly without venture funding.

The focus shifts to David Holtz, founder of Midjourney, who is a respected figure in Silicon Valley. His previous venture, Leap Motion, pioneered mid-air gesture control, though it was criticized for usability issues. Despite these challenges, Holtz's reputation helped him secure resources quickly for Midjourney. This background demonstrates Holtz’s ability to innovate and scale projects, establishing the foundation for the company's success in generative AI.

The paragraph reveals David Holtz's broader goals for Midjourney beyond just creating an image generation tool. The vision is to build new 'human infrastructure' that fosters reflection, imagination, and coordination. Holtz sees Midjourney as a platform for people to reflect on their desires, imagine possibilities, and coordinate to achieve them. Discord’s social layer plays a crucial role in this, creating a collaborative and imaginative environment where users inspire each other.

This paragraph discusses the social aspect of Midjourney’s success, particularly its use of Discord. Initially, Midjourney tested the platform with a bot among its own team, which led to a unique discovery: users become more creative when working in a collaborative environment with strangers. This social interaction leads to an imaginative environment where users build upon each other’s ideas, transforming basic prompts into more elaborate, creative outputs.

The paragraph dives into the technical journey of Midjourney, explaining that the first few models were not self-trained but built by tinkering with open-source components, such as OpenAI’s CLIP. The team trained their own version (version four), which took nine months to develop. A key contributor to this evolution is Katherine Crowson, an independent researcher whose work laid the foundation for diffusion models, a core technology in AI image generation.

This section highlights Midjourney’s efficient use of GPUs around the world, adjusting to nighttime to balance GPU usage across regions. It also points out the challenges of scaling AI models due to the physical and computational limits of current technology. The paragraph concludes with David Holtz’s prediction of two potential futures: one where scaling remains slow and limited, and another where breakthroughs in custom chips could accelerate AI development.

David Holtz discusses the long-term future of AI, where scaling will likely be constrained by computational limits. He outlines two possible scenarios: gradual scaling over seven years or rapid scaling through the development of custom chips that significantly reduce computational requirements. Holtz envisions potential innovations like neural networks embedded directly into chips, which could dramatically reduce energy consumption and increase efficiency.

The final paragraph reflects on the unanswered email to David Holtz, and instead the author pivots to their recent exploration of AI ethics through an interview with ChatGPT. This introduces a broader conversation about the ethical implications of AI-generated art and teases further discussion on the topic, leaving readers intrigued about what’s to come.