Microsoft Discovery Demo: Microsoft Build 2025

Key Concepts:

• Immersion Coolants
• PFAS (Per- and Polyfluoroalkyl Substances) / Forever Chemicals
• Knowledge Graph
• Generative Chemistry
• AI Models
• HPC (High-Performance Computing) Simulations
• Microsoft Discovery
• Iterative Loop (Reasoning, Hypothesis Generation, Experimentation)
• Dielectric Constant
• Boiling Point

1. Introduction and Problem Statement:

• John, the chemistry product lead, introduces the problem of finding environmentally friendly immersion coolants for datacenters.
• The current immersion coolants often rely on PFAS ("forever chemicals"), which are harmful to the environment.
• The goal is to discover a PFAS-free immersion coolant using Microsoft's science platform.

2. Three-Step Iterative Process:

• John outlines a three-step iterative process:
  • Reasoning over Knowledge: Gathering and understanding existing scientific knowledge.
  • Generating Hypotheses: Creating potential solutions based on the gathered knowledge.
  • Running Experiments: Testing the hypotheses and validating the results.

3. Reasoning Over Knowledge:

• The process begins with researching coolants and their properties to identify potential candidates.
• The platform uses a network of agents that reason over scientific knowledge from public sources and internal research.
• A knowledge graph is used to provide deeper insights and more accurate answers compared to LLMs alone.
• The platform generates a summary and a comprehensive report on coolant research, with links to citations for trusted research.

4. Generating Hypotheses:

• The next step involves generating a plan specific to the investigation, informed by the research.
• The plan targets a specific boiling point and a dielectric constant suitable for electronics.
• Microsoft Discovery automatically builds the workflow without requiring the user to specify methods or write code.
• The agents can use tools and models from Microsoft, open-source solutions, third-party solutions, or internal organizational tools.
• The generated plan includes:
  • Generative Chemistry: Creating millions of novel candidates.
  • AI Models: Screening the candidates quickly.
  • HPC Simulations: Validating the findings.

5. Experimentation:

• Microsoft Discovery executes the plan, managing HPC resources in Azure.
• The discovery agents work together in real-time to drive intense computations.
• This process compresses the time to shortlist candidates from months or years to days or hours.

6. Results and Validation:

• The platform identifies a set of candidates for PFAS-free immersion coolants.
• The user can analyze the results to determine if they are ready to proceed to the lab or if another iteration is needed.
• The boiling points and dielectric constants of the candidates are within the desired range.

7. Real-World Application and Discovery:

• The process is not just a demo; a promising candidate was synthesized and tested.
• A video shows a standard PC running Forza Motorsport immersed in the new coolant, maintaining a stable temperature without fans.
• The discovery demonstrates a promising candidate for an immersion coolant that does not rely on forever chemicals.

8. Potential Applications and Conclusion:

• Microsoft Discovery can be used across various domains, including designing new therapeutics, semiconductors, and materials.
• The presentation concludes by encouraging the audience to use Microsoft Discovery for their own breakthroughs.

9. Technical Terms Explained:

• Immersion Coolants: Fluids used to cool electronic components by direct immersion.
• PFAS (Per- and Polyfluoroalkyl Substances): Synthetic chemicals that are persistent in the environment and potentially harmful.
• Knowledge Graph: A structured representation of knowledge that connects entities and their relationships.
• Generative Chemistry: Using computational methods to design and create new molecules with desired properties.
• AI Models: Machine learning algorithms used to predict and screen the properties of molecules.
• HPC (High-Performance Computing): Using powerful computing systems to perform complex simulations and calculations.
• Dielectric Constant: A measure of a material's ability to store electrical energy in an electric field.
• Boiling Point: The temperature at which a liquid changes to a gas.

10. Notable Quotes:

• "These are an interesting line of research for cooling datacenters. Unfortunately, most of them are based on PFAS, or forever chemicals, which are harmful for the environment." - John, highlighting the problem.
• "Behind the scenes, it uses a knowledge graph to provide deeper insights and more accurate answers than we get from LLMs on their own, which can struggle to connect the dots." - John, explaining the advantage of using a knowledge graph.
• "Microsoft Discovery can compress the time to days or even hours." - John, emphasizing the efficiency of the platform.
• "We found a promising candidate for an immersion coolant that does not rely on forever chemicals." - John, announcing the successful discovery.
• "The next great breakthrough is yours to discover." - John, encouraging the audience to use the platform.

11. Synthesis/Conclusion:

The presentation showcases Microsoft Discovery as a powerful platform for accelerating scientific discovery, specifically in the context of finding environmentally friendly immersion coolants. By leveraging a knowledge graph, AI models, and HPC simulations, the platform streamlines the process of reasoning over knowledge, generating hypotheses, and running experiments. The successful discovery of a PFAS-free coolant demonstrates the platform's potential to drive innovation across various scientific domains.