When Will AI Hit the Enterprise? Ben Horowitz and Ali Ghodsi Discuss

This podcast episode explores the challenges faced by enterprises in adopting generative AI, as well as the perspectives on data sharing, building better models for large datasets, potential fragmentation of use cases, and the role of open source models in the AI industry. It discusses the cautious nature of enterprises, their concerns about data privacy and security, and the slow decision-making processes contributing to the slow adoption of generative AI. However, there is a growing realization among CEOs and boards about the potential competitive advantage that generative AI can provide. The episode also highlights the internal discussions and debates within enterprises about sharing their data with external providers. It discusses the importance of building better models with large datasets and the challenges faced by universities in keeping up with the advancements in AI. The role of open source models in shaping and advancing AI research is emphasized, as well as the ongoing competition between open source and proprietary models. The limitations of AI models and the need for human intervention are also addressed, along with ethical considerations and responsible development of AI technologies. Main points • Enterprises face challenges in adopting generative AI, including concerns about data privacy and security, reluctance to share proprietary datasets, fear of data leakage, internal politics, and slow decision-making processes. • There is a growing realization among CEOs and boards about the potential competitive advantage of generative AI, leading to a shift in attitudes towards building their own solutions instead of relying on external providers. • Enterprises are debating the sharing of data with external providers, considering the value and competitive edge it can provide, but also aiming to retain control and ownership of their intellectual property. • Building better models for large datasets involves strategic decision-making, considering the use of existing models, the acquisition of specialized expertise, and the emphasis on specific use cases and accuracy in model training. • The potential fragmentation of use cases in AI highlights the importance of specialization and the need for common base models to underlie multiple applications. • Open source models have played a significant role in shaping AI research, offering accessibility, flexibility, and scalability, while universities face challenges in keeping up with industry resources. • Ethical considerations and responsible development are important in addressing the limitations of AI models, the need for human intervention, and potential risks associated with large models and open-source technologies.