AI-Powered Research: Opportunities and Challenges – Day 2

On February 5, 2025, the “AI-Powered Research: Opportunities and Challenges ” workshop convened at the ELI Beamlines facility in Dolní Břežany. Organized in cooperation with the Institute of Physics and the Programme P4F, the event attracted an international audience of 70 researchers representing 14 different research organizations and universities.

The day’s agenda was structured as follows:

• Morning Session (9:30 – 12:00): Building AI-Enhanced Research Workflows
  • Understanding AI capabilities for research enhancement
  • Responsible AI integration
  • Transitioning from basic AI use to complex research workflows, encompassing stages from research ideation to publication
  • Introducing the concept of an “Expert AI Team” – setting up personalized AI research assistants, including those for grant writing
• Lunch Break and Guided Tour (12:00 – 13:30):
  • Participants enjoyed a 90-minute break, which included lunch and an optional guided tour of the ELI Beamlines facility.
• Afternoon Session (13:30 – 16:30): Practical Demonstration
  • Hands-on AI research workflows – building and customizing AI research teams tailored to specific domains
  • Presentation of 2–3 use cases from MERIT and P4F projects where AI is integrated into the research methodology
• Networking Session (16:30 – 18:00):
  • An opportunity for participants to connect, discuss insights from the day, and explore potential collaborations.

The speaker, Michaela Liegertová, Vice-Dean for Development and Quality at the Faculty of Science, Jan Evangelista Purkyně University in Ústí nad Labem, led an interactive workshop focusing on the practical applications of AI in research. Participants also engaged in practical demonstrations. The session was closed by 2 presentations given by researchers Václav Pavlík and Patrik Čechvala. Pavlík, a MERIT program awardee from the Department of Galaxies at the Astronomical Institute of the Czech Academy of Sciences, and Čechvala, a P4F program awardee from the Department of Astroparticle Physics at the Institute of Physics of the Czech Academy of Sciences, shared insights into how AI is advancing their respective research projects in astronomy and astrophysics.

Few takeaways from the workshop included:

• Choosing the right Model: Aligning the AI model with the specific task is crucial. For instance, models like ChatGPT or Claude are more suitable for text production, while others like STORM, YOU, or Perplexity excel in information search.
• Critical evaluation of AI outputs: AI systems predict rather than think. It is essential to critically assess their outputs, as hallucinations are an inherent feature of large language models (LLMs).
• LLMs and mathematical operations: LLMs may appear to perform mathematical operations, but they are essentially predicting the next token based on learned patterns.
• AI as a thinking partner: Utilizing AI can help explore multiple perspectives and foster out-of-the-box thinking.
• Contextual prompts: Providing ample context in prompts enhances efficiency. Smarter prompts lead to less computation, thereby reducing environmental impact. There are AI tools available to assist in crafting efficient prompts.
• Rapid evolution of AI information: The information shared during the workshop is subject to rapid change, potentially becoming outdated within weeks.