Chats, a new series from ‪@johnwileysons‬ explores the practical challenges and opportunities facing academic publishing and research.

Episode One | The AI Choice: Why Waiting Isn't an Option

Choosing, Training and Using AI Models: Expert Interview with BMJ's CTO, Ian Mulvany and Ray Abruzzi (Senior Director of AI Product Management at Wiley). They explore the intersection of #ai development and academic publishing, offering insights for developers building research tools, publishers evaluating AI partnerships, researchers adopting new technologies, and institutions making strategic AI investments.

They discuss the importance of subject matter expert integration in AI development, fine-tuning strategies, prompt engineering, considerations for publisher-developer partnerships, model performance risk assessment, misconceptions around hallucinations and best practices, and more.

👉 Learn more about Wiley’s content licensing opportunities for AI developers: https://bit.ly/46k8HfW

Key Topics Covered:

• Fine-tuning AI models, prompt engineering, and deployment strategies
  
  
• Evaluating AI tools and building transparent publisher-developer partnerships
  
  
• Practical tips for researchers on choosing and using #aitools effectively
  
  
• Addressing misconceptions about #aihallucinations and model limitations
  
  

For AI developers:

• Statistical inference vs. retrieval: why LLMs aren't search engines
  
  
• Fine-tuning strategies, ensemble methods, and context window optimization
  
  
• Prompt engineering insights: why language mastery outperforms technical precision
  
  
• Production deployment lessons from high-stakes environments
  
  

For publishers & institutions:

• Considerations for AI tool evaluation and selection.
  
  
• Building partnerships with AI companies and requesting transparency
  
  
• The publisher's role in AI's "truth infrastructure" and moral obligations
  
  
• Investment priorities: where untapped AI potential lies
  
  

For researchers:

• Practical guidance on choosing reliable AI tools for academic work
  
  
• Understanding AI limitations: when to trust outputs and when to verify
  
  
• Prompt engineering techniques for better research outcomes
  
  
• Avoiding the "search engine misconception" that leads researchers astray
  
  
• Risk assessment for different use cases