Can o3-pro write a book? ElevenLabs v3 and MCP as the AI Business Model - EP99.08-PRO

Chris and Mike delve into the advancements of Eleven Labs V3, particularly its AI podcast feature that incorporates emotional tags into voice recordings. While Mike questions the effectiveness of voice training, Chris highlights the technology's potential for creating realistic emotional expressions, especially in performances like Shakespearean plays. They discuss the accessibility of Eleven Labs' models and improvements in voice quality compared to others.

The conversation shifts to AI models, including OpenAI's O3 and Gemini 2.5 Pro, noting recent price drops that enhance O3's competitiveness. They speculate on OpenAI's pricing strategy and emphasize the importance of speed and availability in AI tools. They reflect on O3 Pro's new pricing structure, which makes it more accessible for testing, and discuss the need for users to learn effective prompt crafting. They acknowledge that while O3 Pro can provide precise answers, it may still produce incorrect information.

The discussion includes the Model Control Protocol (MCP) and two approaches to integrating MCP servers: public and self-hosted. The self-hosted option offers greater control and insight into the model's decision-making process. Emerging techniques like asynchronous tool calling in models such as Claude 4 are noted for their superiority in retrieving information.

The speakers critique OpenAI for not providing more accessible solutions and advocate for user-friendly hosting for MCPs, predicting a future where simpler installations for tools will be prevalent. They envision AI assistants evolving to perform tasks autonomously with user approval, emphasizing the need for effective information retrieval methods.

A recent incident involving a Boeing 787 Dreamliner crash underscores the importance of accurate information gathering. The speakers highlight AI's effectiveness in research, noting that it can provide broad analyses and challenge traditional journalism's reliance on social media. They discuss the potential for curated sources to represent a new business model for data and journalism, moving away from traditional article writing.

The conversation touches on digitizing old manuscripts and creating searchable databases, proposing subscription services for accessing proprietary journals through AI. They anticipate a competitive rush for these services, with users seeking subscriptions for curated knowledge. An experiment is introduced to evaluate the creative writing abilities of three AI models tasked with writing a space-themed version of "The Count of Monte Cristo," with subjective critiques on their outputs.

The effectiveness of different AI models is further explored, particularly the transition from Sonnet 3.5 to Sonnet 4, with critiques on the latter's performance. Excitement surrounds O3 Pro for its high-quality responses and tool-calling abilities, with plans to test it on real-world tasks. The discussion concludes with anticipation for future developments in LLMs and MCPs, focusing on the orchestration of tools rather than the superiority of chat models.