Microsoft has launched a new platform for enterprises that uses artificial intelligence to accelerate scientific research and developement. This could compress years of laboratory work in weeks or days.
This platform, called Microsoft Discoveryleverages specialized AI and high-performance computers to help scientists and engineering tackle complex research challenges, without requiring them write code, the firm announced Monday at its Build developer conference.
In an exclusive interview with VentureBeat, Microsoft’s Corporate Vice President for Strategic Missions and Technologies, Jason Zander, said, “We are looking at how we can apply advances in agentic AI and computing work, and then onto quantum computing, and use it in the most important space, which in science.”
Microsoft’s research has already shown the potential of this system, which helped discover a novel cooling agent for immersion cooling in data centers in just 200 hours. This process would normally have taken months or even years.
Zander said that in 200 hours, the framework allowed him to screen 367,000 candidates he had come up with. “We took it to our partner and they synthesized it.”
How Microsoft is putting supercomputing in the hands everyday scientists
Microsoft Discovery is a significant step towards democratizing advanced scientific software, allowing researchers the ability to interact with supercomputers or complex simulations without requiring specialized programming knowledge.
Zander said, “It is about empowering scientists with agentic AI to transform the entire process of discovery.” “My PhD is biology.” I’m no computer scientist, but it’s very powerful if you let me control a supercomputer by simply prompting it.
This platform addresses a major challenge in scientific research – the disconnect between domain knowledge and computational skills. Scientists have traditionally had to learn programming in order to use advanced computing tools. This created a bottleneck for the research process. This democratization is especially valuable for smaller research institutes that lack the resources to hire computational experts to augment their scientific team. Microsoft has lowered the barrier to entry to cutting-edge research by allowing domain experts and scientists to use natural language to query complex simulations.
As a scientist, i’m a biologist. I don’t understand computer code. Zander said, “I don’t want spend all my time writing scripts in an editor to ask a computer to do something.” “I wanted to say, ‘This is what I want, in plain English, or plain language. Go do it.'”
Inside Microsoft Discovery: AI “postdocs” that can screen hundreds and thousands of experiments
Microsoft Discovery is operated by what Zander called a team AI “postdocs”specialized agents who can perform different aspects in the scientific process from literature review to computation simulations.
Zander explained that “these postdoc agents perform this work.” It’s like a team of people who just got their PhD. They’re like medical residents — you’re working in the hospital but still finishing.
This platform combines two key components – foundational models for planning and specialized models for specific scientific domains such as physics, biology, and chemistry. This approach is unique in that it combines general AI capabilities with deep specialized scientific knowledge.
Zander said, “The core process is divided into two parts.” “We’re using the foundational models to do the planning,” Zander said. The AI side has a set models that are specifically designed for certain domains of science. This includes physics chemistry and biology. This allows it understand conflicting theories, diverse experimental results and maintain transparency by tracking the sources and reasoning processes.
The Copilot interface is at the heart of the user’s experience. It orchestrates these specialized agent based on the researcher’s prompts. It identifies which agents to leverage, and sets up end-toend workflows. This interface is a central hub for human scientists to guide their virtual research team.
Microsoft’s AI helped solve a cooling problem in a data center that took months to solve
Microsoft Discovery addresses a pressing issue in data center technology – finding alternatives to coolants that contain PFAS (so-called “forever chemical” substances) which are increasingly subject to regulatory restrictions.
The cooling methods used in data centers are often based on harmful chemicals, which are becoming unsustainable as global regulations move to ban them. Microsoft researchers used this platform to screen hundreds and thousands of possible alternatives.
We did prototypes for this. Zander stated that he had done a prototype when he owned Azure eight years ago. It worked really well. “It is actually 60 to 90 percent more efficient than air cooling.” The coolant material on the market contains PFAS.
Microsoft synthesized a coolant and tested it cooling a GPU that was running a videogame. This application is still experimental, but it shows how Microsoft Discovery is a tool that can help companies with regulatory challenges to reduce the time it takes to develop new products.
Microsoft’s data centers are not the only ones affected by this. This approach could be used by any industry that is under similar regulatory pressures to replace existing chemicals or materials. What would have taken years to develop could now be completed within months.
Daniel Pope is quoted in a press release as saying: “The speed of molecular screen achieved by Microsoft Discovery was impossible with traditional methods.” Microsoft Discovery can do in weeks what used to take years of lab work, trial-and error, and confidence.
Microsoft’s new AI platform is being used by major companies in the fields of pharma, beauty and chips.
Microsoft has built an ecosystem of partners from diverse industries to implement it, indicating that its applicability goes beyond the company’s own research needs.
Pharma giant GSK explores the platform’s potential to transform medicinal chemical. The company announced its intention to partner with Microsoft in order to advance “GSK’s generative platforms for simultaneous prediction and testing to create new medicines with greater speeds and efficiency.”
Estee Lauder (19459051) plans to use Microsoft Discovery to accelerate the development of products in skincare, make-up, and fragrance. Kosmas Kratsos, PhD MBA, Vice President for R&D and Innovation Technology, Estee Lauder Companies, said that the Microsoft Discovery platform would help them to unlock the power of their data to drive rapid, agile, breakthrough innovations and high-quality personalized products. Microsoft will also expand its partnership with Nvidia by integrating Nvidia’s technology. ALCHEMI Microsoft Discovery combines BioNeMo Microservices to enable faster breakthroughs in materials science and life sciences. This partnership will enable researchers to leverage the latest inference capabilities for candidate detection, property mapping, or synthetic data generation.
Dion Harris, Nvidia’s senior director of accelerated Data Center Solutions, said that AI is accelerating scientific discovery. Microsoft will integrate Synopsys industry solutions in the semiconductor space to accelerate chip development and design. Sassine Gahzi, President and CEO at Synopsys, said that semiconductor engineering is “among the most complicated, consequential, and high-stakes science endeavors of our times,” making it an “extremely compelling use case for AI.”
Microsoft’s quantum strategy – Why Discovery is only the beginning of the scientific computing revolution?
Microsoft Discovery represents a steppingstone towards the company’s larger quantum computing ambitions. Zander explained that the platform uses high-performance conventional computing but is designed with future quantum capabilities as a priority. Zander stated that “Science is the hero scenario of a quantum computing system.” “If you ask yourself what can a computer do?” It is extremely good at exploring complex problem spaces that traditional computers are not able to do.
Microsoft announced recent advancements in quantum computer with its Zander explained that “general generative chemistry” is the “hero scenario” for high-scale quantum computing. It can take a small amount and explore a space which would take millions of year for a supercomputer or a classic computer to do.
Zander imagines a future in which quantum computers will design their own successors. “One of the things I want to do as soon as I get a quantum computer that can do that kind of work, is to give it my material pile for my chip.” I’m basically going to say, “Okay, go simulating that sucker.” Tell me how to build a newer, better version of yourself. ‘”
Microsoft’s scientific platform has built-in ethical safeguards to prevent misuse
Given the powerful capabilities of Microsoft Discovery, questions about misuse are a given. Zander stressed that the platform incorporates Microsoft’s responsible AI framework. Zander stated that “we have the Responsible AI program and it has been around for a while, in fact, I think we were the first company to actually put this kind of framework in place,” Zander. “Discovery is absolutely following all responsible AI guidelines.” The company seems to be taking an active approach to identifying possible misuse scenarios. Zander explained that “we already look for certain types of algorithms which could be harmful, and try to flag them in a content moderation style.” “Again, this analogy would be similar to what a bot for consumers would do.”
The focus on responsible innovation reflects that powerful scientific tools can be used in multiple contexts. For example, the same platform which could speed up lifesaving drug discoveries could be misused elsewhere. Microsoft’s approach aims to balance innovation and appropriate safeguards. However, the effectiveness of the measures will only be apparent as the platform becomes more widely adopted.
The bigger picture: Microsoft’s AI platform can reshape human innovation
Microsoft has entered the field of scientific AI a t a time of rapid discovery. The ability to compress the research timelines can have profound implications in addressing urgent global issues, from drug development to climate change solutions.
Microsoft’s approach differs in its focus on accessibility and integration with its existing cloud infrastructure, as well as its future quantum ambitions. Microsoft could remove a significant bottleneck to scientific progress by allowing domain experts directly to leverage advanced computing, without intermediaries.
Zander explained that “the big efficiencies come from places where instead of me cramming extra domain knowledge, in this instance, a scientist who has learned to code, are saying, “Actually, let the genetic AI handle that, and you can do what your PhD allows you to do, which is get forward progress.”
The democratization and accessibility of advanced computational methods may lead to a fundamental change in the way scientific research is conducted worldwide. Smaller labs in regions with less computing infrastructure could suddenly have access to capabilities that were previously only available to elite research institutions.
The success of Microsoft Discovery depends on how well it integrates with complex existing research workflows, and whether its AI agents are able to truly understand the nuances in specialized scientific domains. Microsoft will have to show consistent, reproducible results in order to gain widespread acceptance. The scientific community is notoriously skeptical and rigorous about new methodologies.
Today, the platform will enter private preview. Pricing details are yet to be announced. Microsoft says that smaller research labs can access the platform via Azure. The costs will be structured similar to other cloud services.
Zander stated that the goal of Microsoft’s business is to enable the core platform rather than having you stand up. “It will just ride on top the cloud and make things much easier for people.”
Accelerating the Future: When AI meets Scientific Method
Microsoft is at a unique point in the history of computing and scientific discovery as it builds out its ambitious AI platform. The scientific method, a centuries-old process, is now being enhanced by some of most advanced artificial intelligences ever created.
Microsoft Discovery is a bet on the fact that the next generation of scientific breakthroughs will not come from brilliant human minds working alone, but rather from the collaboration between them. AI will handle the computational heavy-lifting, while humans provide the creativity and intuition that machines lack.
Zander said, “If you think of chemistry and materials sciences, materials have a direct impact on 98% of the planet.” “Everything. The desks, displays, clothing, and the clothing we wear. It’s all material.”
The implications for accelerating discovery in these areas extend far beyond Microsoft or the tech industry. Microsoft Discovery, if successful, could fundamentally change the pace at which humanity innovates in response to existential issues – from climate changes to pandemic prevention.
It’s not a question of whether AI will transform science research, but rather how quickly and deeply. Microsoft believes that combining human scientific expertise with agentic AI could be the answer to a world faced with increasingly complex challenges.
Daily insights into business use cases from VB Daily
Want to impress your boss? VB Daily can help. We provide you with the inside scoop about what companies are doing to maximize ROI, from regulatory changes to practical deployments.
Read our privacy policy
Thank you for subscribing. Click here to view more VB Newsletters.
An error occured.
