Microsoft aims to harness AI to accelerate scientific breakthroughs across fields like chemistry, biology, and materials science with new research tools
Is it possible for AI to expedite certain aspects of the scientific process? Microsoft seems to concur.
According to a press release provided to TechCrunch, Microsoft Discovery, a platform that utilizes agentic AI to “transform the [scientific] discovery process,” was unveiled at the company’s Build 2025 conference on Monday. Microsoft asserts that Microsoft Discovery is “extensible” and capable of managing specific science-related duties “end-to-end.”
“Microsoft Discovery is an enterprise agentic platform that accelerates research and discovery by transforming the entire discovery process with agentic AI, including scientific knowledge reasoning, hypothesis formulation, candidate generation, and simulation and analysis,” Microsoft explains in its release.
“The platform facilitates the collaboration between scientists and researchers and a team of specialized AI agents to facilitate the rapid, scaled, and precise advancement of scientific outcomes through the application of the most recent advancements in AI and supercomputing.”
Microsoft is one of numerous AI laboratories that are optimistic about the potential of AI to advance science. Google introduced an “AI co-scientist” earlier this year, which the technology company claimed could assist scientists in formulating hypotheses and research plans.
Anthropic and its primary competitor, OpenAI, as well as institutions such as FutureHouse and Lila Sciences, have claimed that artificial intelligence (AI) tools have the potential to significantly expedite scientific discovery, particularly in the field of medicine.
However, many researchers do not believe that AI is particularly beneficial in directing the scientific process, primarily because of its unreliability.
Anticipating an infinite number of perplexing variables is one of the obstacles to cultivating an “AI scientist.” It is uncertain whether AI is capable of unconventional problem-solving that results in genuine breakthroughs, such as narrowing down a vast list of possibilities. However, it may prove advantageous in situations where extensive exploration is required.
The results of AI systems specifically designed for scientific applications have been largely disappointing thus far.
In 2023, Google announced that approximately 40 novel materials had been synthesized with the assistance of one of its AIs, GNoME. However, an external analysis revealed that none of those materials were, in fact, novel. In the interim, numerous organizations that utilize artificial intelligence (AI) for drug discovery, such as BenevolentAI and Exscientia, have encountered significant clinical trial failures.
Microsoft undoubtedly anticipates that its endeavor will surpass those that have preceded it.
