The AI Revolution Will Not Be Monopolized: How open-source beats economies of scale, even for LLMs

Open source models and software provide key advantages through interoperability, transparency, and ability to customize/extend functionality rather than just being free

The AI industry can be divided into two main categories:

• Human-facing systems (like ChatGPT, customer-facing products)
• Machine-facing models (underlying technology components)

Large tech companies don’t gain lasting monopolies through data alone:

• User data helps improve products but doesn’t guarantee market dominance
• Core AI technology is based on published research
• Competition exists through open source alternatives

Task-specific models offer advantages over large general models:

• Faster and cheaper to run
• More predictable and controllable
• Can be fine-tuned with limited data
• Better suited for specific business use cases

Successful AI implementation often combines:

• Using APIs/larger models during development
• Deploying smaller, specialized models in production
• Custom training on domain-specific data
• Modular, swappable components

Companies compete primarily on:

• Product features and user experience
• Customization capabilities
• Integration and deployment options
• Price and performance
• Not the underlying AI technology itself

Regulation should focus on:

• Products/applications rather than base technology
• Preventing artificial monopolies
• Maintaining open competition
• Distinguishing between human-facing and machine-facing systems