Why machines don't speak biology: Toward native biological language models

Area of Science:

• Computational Biology
• Artificial Intelligence
• Systems Biology

Background:

• Current artificial intelligence (AI) research aims for universal biological models.
• Living systems are complex, evolved, and emergent, making their behaviors difficult to predict solely from component parts.

Purpose of the Study:

• To propose a novel framework for grounding AI in fundamental biological processes.
• To develop data-driven biological world models with explicit mechanistic connections.

Main Methods:

• Integrating AI with canonical biological processes.
• Constructing world models that link molecular and cellular levels.
• Incorporating spatial and temporal dynamics into biological models.

Main Results:

• A proposed methodology for building biologically grounded AI.
• Framework for understanding emergent behaviors from mechanistic links.
• Foundation for data-driven models across biological scales.

Conclusions:

• Grounding AI in canonical biological processes is essential for accurate biological modeling.
• Explicit mechanistic links across scales are key to capturing emergent biological phenomena.
• This approach offers a path toward more robust and predictive AI for biology.