'My Cofounder Stole My Idea' — An Entrepreneur's Wild Journey To Build An $83 Million AI Company

Key Concepts

• AI Creative Suite: Software that uses artificial intelligence to assist and enhance creative processes.
• Image Generation: AI models that create images from text prompts.
• Video Generation: AI models that create videos from text prompts or other inputs.
• Enhancer/Upscaler: AI tools that improve the quality, resolution, or detail of existing images or videos.
• Real-time Tool: AI tools that provide immediate feedback or generation, often integrated with other creative software.
• Deep Learning: A subset of machine learning that uses artificial neural networks with multiple layers to learn from data.
• Generative AI: AI models capable of creating new content, such as text, images, audio, and video.
• Fine-tuning (LoRA): A process of adapting a pre-trained AI model to a specific task or dataset, often by training only a small portion of the model.
• Data Set Curation: The process of selecting, organizing, and preparing data for training AI models.
• Controllability: The ability of a user to precisely guide and influence the output of an AI model.
• Aesthetics: The principles concerned with the nature and appreciation of beauty, especially in art.
• Vertical Integration: A strategy where a company controls multiple stages of its production process, from raw materials to final product.
• Horizontal Platform: A platform that offers a wide range of services or functionalities across different areas.
• LLMs (Large Language Models): AI models trained on vast amounts of text data, capable of understanding and generating human-like text.
• Multimodal Architectures: AI systems that can process and integrate information from multiple types of data (e.g., text, images, audio).
• Sustainability in AI: The environmental impact of AI, including energy consumption and resource usage.

Korea: Empowering Creatives with AI

Victor Perez, co-founder and CEO of Korea, an AI creative suite backed by $83 million, describes Korea as a platform that brings the latest AI research to creatives through intuitive tools. The most popular tools include image generation, video generation, an enhancer, and a real-time tool.

Perez's Creative Journey and the Genesis of Korea

Perez's creative background began with classical guitar, leading to a passion for music and eventually photography and graphic design for his band. He pursued Visual Systems Engineering, a blend of music and computer science, which led him to coding and AI. This interdisciplinary path merged his interests, sparking his exploration of creative AI applications.

He was inspired by the rapid advancements in deep learning around 2017, noting that academic curricula lagged behind the cutting edge. This gap, coupled with a realization that many AI tools were geared towards programmers rather than artists, motivated him to build more accessible interfaces.

The inception of Korea was serendipitous. Perez began building side projects to create user-friendly interfaces for AI, not with the initial intention of starting a company. One of these projects attracted the attention of venture capital firm Age of Zero. In a chaotic turn of events, his co-founder, Diego, pitched Perez's project during an interview for a fellowship, leading to an invitation to Miami and ultimately the formation of Korea in late 2021/early 2022.

Evolving Use Cases and Korea's Impact

Since 2022, AI capabilities have advanced significantly. Perez observes that AI has transitioned from a novelty to a practical tool for creatives. Current use cases include:

• Idea Generation and Mood Boards: Creatives use AI to generate multiple visual concepts quickly, serving as a more dynamic alternative to platforms like Pinterest.
• Enabling New Creatives: AI empowers individuals who may not identify as artists to express their ideas visually.

For professional creatives, Korea offers efficiency:

• Architecture Studios: Utilize the enhancer tool to upscale basic 3D renders into high-detail 4K images in minutes, a process that previously took hours and was costly.
• Industrial Designers: Employ the real-time tool to connect with 3D software like Blender or Cinema 4D, enabling rapid visualization of product designs (e.g., chairs, helmets) from simple shapes and doodles.

For hobbyists and non-professional creatives, Korea provides an enabling experience, allowing them to visualize their ideas for fun, akin to the appeal of photography.

The Technology Behind Korea

Korea utilizes a diverse array of AI models, with approximately 20-30 different models, each tailored for specific use cases or workflows. The core principle behind these models is their ability to learn patterns from vast amounts of data (images, videos). This learning allows them to generalize, enabling the creation of novel combinations, such as an avocado-shaped chair, even if such an object has never been seen before. This generalization is key to the creative potential of these AI models.

AI Trainer: A Surprising Success

A notable example of user-driven innovation was the "AI Trainer" feature. Initially perceived as too complex for creatives, the hardcore AI community began curating datasets and fine-tuning models to achieve specific styles and controllability. This demand led Korea to develop "Create a Train," allowing users to upload images to train AI on specific styles, people, or objects. This feature became one of Korea's most successful launches, highlighting creatives' desire for controllability and specific aesthetic outcomes.

Enterprise Adoption and Branding Capabilities

Major enterprises like Perplexity, Pixar, and Shopify are leveraging Korea, particularly its AI training capabilities. These companies use fine-tuned models (LoRAs) for specific campaigns, enabling them to generate highly polished, AI-driven content that often appears indistinguishable from human-created work. Enterprise clients demonstrate a high level of knowledge about AI, often inquiring about model training capabilities upon joining the platform.

Implementing Taste and Human Creativity

Korea's existence is rooted in the idea that "taste needs a new form of IP." The platform aims to provide tools for creatives to imbue AI with their specific taste and ideas. This is achieved through the process of training models, where users curate datasets that represent their aesthetic preferences. Achieving desired results often involves iterative refinement, mirroring the human creative process of drafting and iteration. This emphasizes that creating with AI is a workflow, not just a simple prompt-to-image generation.

The Future of Creatives: Shift, Not Replacement

Perez views the conversation around AI replacing artists as evolving, with a growing openness to adoption. He likens the shift to the transition from hand-drawn logos to computer-generated ones, or from painted portraits to photography. AI is seen as a medium that will lead to new, powerful formats requiring creatives to possess a deep understanding of AI.

While some express skepticism or fear, Perez notes that most creatives he speaks with are excited, viewing AI as an empowering tool. He believes adaptation is key, and that AI offers greater possibilities for creating formats previously unimaginable. This new landscape requires a new skill set, emphasizing taste and ideas, with parallels drawn to coding where AI is more advanced but faces less pushback.

The future may see job postings for "AI Creatives" or AI proficiency becoming a standard requirement for creative roles. While AI can prototype and generate content, achieving production-level results still requires human oversight and a refined creative process.

Investor Confidence and Korea's Strategy

Korea's $83 million funding reflects investor confidence in its strategy of prioritizing product development and user experience before building its own AI infrastructure and training models. This "horizontal platform" approach, combined with vertical integration, appeals to investors. Their focus on controllability and aesthetics for creative workflows is a key area of development.

Sustainability in AI

While Korea is not currently using an "insane amount of compute," Perez acknowledges the sustainability concerns surrounding AI, particularly for LLMs. He believes the potential benefits of AI in areas like medical research (e.g., curing cancer) or developing sustainable materials and optimized infrastructure could justify the energy expenditure. The ultimate value depends on AI's ability to drive significant positive impact.

The Next 1-5 Years: Advancing Media Models

Perez is most excited about the rapid progress expected in media models (video, image, audio, 3D), which he believes are currently behind LLMs in terms of performance. He anticipates advancements in reasoning capabilities and multimodal architectures, leading to improved communication between artists and models, and significantly enhanced model capabilities. He envisions video models potentially offering 3D experiences and new ways of interacting with generated content. Korea aims to be a research lab exploring these emerging capabilities and translating them into intuitive products for the creative community.