This article by Wharton professor Kartik Hosanagar and Ramayya Krishnan, Dean of the Heinz College of Information Systems and Public Policy at Carnegie Mellon University, delves into the evolving landscape of generative AI technology. The article explores the key players, opportunities, and challenges within this field. It highlights the importance of understanding the value chain, making decisions between building and borrowing AI models, and navigating emerging challenges for success. The article also discusses the various components of the generative AI ecosystem, such as the hardware layer, data, computing infrastructure, foundation models, RAGs and fine-tuned models, and LLM applications. It emphasizes the battle for value in the foundation model market, the choices companies must make in building or borrowing models, and the advantage that organizations with domain-specific data have. The article concludes with key takeaways for managers, including leveraging domain expertise, integrating AI into existing products, and carefully evaluating the use cases of generative AI. Finally, the article highlights copyright issues as an emerging challenge that may favor established players.
| Signal | Change | 10y horizon | Driving force |
|---|---|---|---|
| Generative AI landscape exploration | Understanding value, challenges in AI | Consolidation of foundation model market | High barriers to entry, demand-side network effects |
| Importance of hardware, data, computing | Foundation of AI ecosystem | Advancements in GPU technology | Increasing processing power, improved datasets |
| Dominance of foundation models | Potential for consolidation | Emergence of a few giant players | High barriers to entry, demand-side network effects |
| Choices between building vs borrowing | Decisions for companies entering the market | Increase in open-source model usage | Convenience, cost-effectiveness, security concerns |
| Domain-specific advantage | Specialized models outperforming generics | Differentiation based on user interfaces | Access to large, high-quality data in specific domains |
| Key takeaways for managers | Strategies for incumbents and entrepreneurs | Integration of AI into existing products | Domain expertise, agility, exploiting inertia |
| Copyright issues in training models | Legal challenges for using copyrighted data | Favoring established players in legal battles | Resources to navigate legal battles, licensing agreements |