Microsoft's new AI platform to revolutionize scientific research—Know how
What's the story
Microsoft has unveiled a new platform, called Microsoft Discovery, at its Build 2025 conference.
The tech giant touts the innovative tool leverages the power of agentic AI to "transform the [scientific] discovery process."
The platform, according to Microsoft, is "extensible" and can handle certain science-related workloads "end-to-end."
The move comes as other tech companies also explore AI's potential in speeding up scientific discovery.
Platform features
Microsoft Discovery: A comprehensive platform for researchers
Microsoft describes its new platform as an "enterprise agentic platform" to expedite research and discovery.
The company says it transforms the entire discovery process with agentic AI, from scientific knowledge reasoning to hypothesis formulation, candidate generation, simulation, and analysis.
The platform aims to enable scientists and researchers to collaborate with specialized AI agents for faster, scalable, and more accurate scientific outcomes using advanced AI and supercomputing technologies.
Industry trends
Other tech giants exploring AI's potential in scientific discovery
Microsoft isn't the only one exploring AI's potential in scientific discovery.
Earlier this year, Google had also brought an "AI co-scientist" to help scientists with hypothesis creation and research planning.
Other companies like Anthropic, OpenAI, FutureHouse, and Lila Sciences have also been trying out AI tools to speed up scientific discovery processes, especially in the field of medicine.
Reliability issues
Concerns persist over AI's reliability in scientific process
Despite the potential benefits, many researchers remain skeptical about AI's current utility in guiding the scientific process, owing to its unreliability.
A major challenge in creating an "AI scientist" is predicting countless confounding factors.
While AI may be useful for broad exploration tasks (like narrowing down extensive lists of possibilities), its ability to perform out-of-the-box problem-solving that leads to genuine breakthroughs remains uncertain.
Performance challenges
Mixed results from AI systems designed for scientific research
So far, the performance of AI systems specifically designed for scientific research has been largely disappointing.
In 2023, Google claimed that it had synthesized around 40 new materials using one of its AIs, GNoME. However, an external review found that not a single one of them was genuinely new.
Several companies using AI for drug discovery have also faced significant clinical trial failures.