What is Microsoft’s approach to AI?

A graphic showing lines and dots in the background behind four photographs representing the four areas of Microsoft's approach to AI. A man in the lower left is shown controlling a drone flying in the upper right corner.

At Microsoft, we believe artificial intelligence (AI) is the defining technology of our time.

We have been on the forefront of cutting-edge research in AI and integrating these powerful, innovative AI technologies into our products and services to help customers do more. Microsoft AI, powered by Azure, provides billions of intelligent experiences every day in Windows, Xbox, Microsoft 365, Teams, Azure AI, Power Platform, Dynamics 365 and Microsoft Defender.

Our AI tools and technologies are designed to benefit everyone at every level in every organization. They are used in workplaces, home offices, academic institutions, research labs and manufacturing facilities around the world, and they are helping everyone from scientists and salespeople to farmers, software developers and security practitioners.

We have made huge investments in AI because we are optimistic about what it can do to help people, industry and society, and because we’re committed to bringing technology and people together to realize the promises of AI responsibly.

To learn more about Microsoft’s work with AI, read about:

Our approach to using AI responsibly

Microsoft believes that when you create powerful technologies, you also must ensure that the technology is developed and used responsibly. We are committed to a practice of responsible AI by design, guided by a core set of principles: fairness, reliability and safety, privacy and security, inclusiveness, transparency and accountability.

Microsoft is putting these principles into practice across the company to develop and deploy AI that will have a positive impact on society.

“With the right guardrails, cutting-edge technology can be safely introduced to the world to help people be more productive and go on to solve some of our most pressing societal problems,” says Natasha Crampton, the chief responsible AI officer at Microsoft.

AI systems are the product of many different decisions made by those who develop and deploy them. From system purpose to how people interact with AI systems, we need to guide these decisions toward beneficial and equitable outcomes.

“That’s what our practice of responsible AI by design is all about,” Crampton says. “We ensure that responsible AI considerations are addressed at the earliest stages of system design and then throughout the whole lifecycle, so that the appropriate controls and mitigations are baked into the system being built, not bolted on at the end.”

This approach does not eliminate all risks, and a commitment to listening, learning and improving is paramount. But it does encourage developers to be clear about any limitations, account for intended uses and potential misuses, and think expansively about how to secure the benefits of a system and guard against its risks.

We believe proactive, self-regulatory efforts by responsible companies help pave the way for these new laws, but we recognize that not all organizations will adopt responsible practices voluntarily.

Microsoft President, Brad Smith, recently outlined the importance of the company stepping up to meet the current AI moment, including calling for thoughtful policy.

“Countries and communities will need to use democratic law-making processes to engage in whole-of-society conversations about where the lines should be drawn to ensure that people have protection under the law,” Smith wrote.

“Effective AI regulations should center on the highest risk applications and be outcomes-focused and durable in the face of rapidly advancing technologies and changing societal expectations. To spread the benefits of AI as broadly as possible, regulatory approaches around the globe will need to be interoperable and adaptive, just like AI itself,” he added.

Microsoft believes democratic, law-making processes are a vital part of a global dialogue with industry, academia and civil society to create principled and actionable norms that ensure organizations develop and deploy AI responsibly.

As part of this process, we are committed to publicly sharing the company’s learnings and best practices, along with the tools that guide our efforts. This includes our Responsible AI Standard, a framework for translating our high-level principles into actionable guidance for our engineering teams.

Like all transformative technologies, we are aware that AI has its risks. Some people will use this technology to exploit the flaws in human nature, target people with false information, undermine democracy and cause harm. We need to plan for and mitigate these risks.

Microsoft has been working on its cross-company, cross-discipline, responsible AI effort for more than six years, creating a strong foundation upon which to keep building for the future. We are clear-eyed about the need to continue moving forward with urgency to keep pace with the rapid evolution of technology and changing societal expectations.

“Fundamentally, our AI work is grounded in our company mission to help every person and organization on the planet to achieve more,” says Crampton, “and it is undergirded by our steadfast commitment to the responsible development and use of AI.”

Our approach to AI research

For more than 30 years, Microsoft Research has been advancing the foundations of computing and translating new scientific understanding into innovative technologies to create value for our customers and broad benefit to society.  

Our researchers collaborate across disciplines, institutions and geographies to deliver cutting-edge advances in vision, speech, language, decision-making and machine learning. They have pioneered AI breakthroughs in conversational speech recognition, machine translation, image captioning, natural language understanding and commonsense question answering

Recent efforts have focused on developing large-scale models that can process information in increasingly sophisticated ways while also becoming more natural and intuitive to use. Advances in deep learning, coupled with internet-scale datasets and Microsoft Azure’s increasingly powerful AI supercomputing resources, have made it possible to create AI models that perform a broad range of tasks across many different applications.   

“Large-scale AI is shifting the landscape of computing research,” says Ashley Llorens, vice president and managing director at Microsoft Research. “As we orient around that shift, you’ll see new frontiers that advance our understanding of human and machine intelligence and how they can intersect and reinforce each other in profound new ways.” 

Our research has played a key role in evolving model architectures and creating AI technology that is more efficient and more adaptable across an even broader range of tasks.

Microsoft researchers have been working on these problems for years, developing expertise in areas like parallel computation that allows people to more quickly train machine learning models at unprecedented scale. This has led to innovations like DeepSpeed, an open-source, deep learning optimization library for distributed training that was developed by Microsoft Research and now is used by the broader computing community.  

We also are focused on delivering value and solving real-world problems through Microsoft products and services. Project FarmVibes, for example, merges AI and data to offer open-source tools that can help farmers adopt sustainable agriculture practices. And the Microsoft Climate Research Initiative provides our research and computing capabilities to a team of multidisciplinary scientists working together to address climate change.   

Our new AI4Science organization is focused on applying deep learning to the natural sciences to model and predict natural phenomena and help address critical issues such as climate change, green energy and pharmaceutical discovery. Our project teams, which include a global collective of researchers and engineers, are exploring a new approach to machine learning, generating training data by simulating natural phenomena from fundamental equations rather than using datasets from the internet.   

This could enable researchers to understand and predict natural phenomena at scales ranging from quantum to galactic and, in turn, drive breakthroughs such as the discovery of new materials that can remove carbon from the atmosphere, Llorens says.    

“At the end of the day, we’re focused on pushing the frontiers in ways that enhance the human experience and positively impact society as a whole,” he says.   

“To achieve this, we’ll need to engage as part of the global research community and continually challenge our assumptions about what is possible. That’s what will produce the future advancements we need.”   

Our approach to AI infrastructure 

With growing confidence in AI and businesses aiming to do more with less, customers are looking for a trusted partner to streamline adoption and rapidly apply intelligence across workloads to improve operations, drive efficiencies and reduce costs. More than a decade ago, we forecast this exponential growth in demand for AI systems and started to build special computing infrastructure to handle it.   

Today, Microsoft’s AI platform, Azure AI, offers infrastructure optimized and purposely built for running large AI models that are ushering in a new era of productivity and creativity. Thanks to our investments, we’re able to deliver a wide range of AI-powered products that fit the needs of our customers and also deliver best-in-class performance and scale for the most compute-intensive AI training and inference workloads.   

Our unique architecture design combines the fastest graphics processing units, or GPUs, available in the market along with a network architecture that chains together thousands of GPUs to enable AI model training and inference at scale.  

“Having thousands of GPUs with high-bandwidth interconnect enables everything else from there,” says Eric Boyd, Microsoft corporate vice president for AI Platform.  

Organizations large and small are developing Azure AI solutions because they can achieve more at scale, more easily, with the proper enterprise-level privacy, security and responsible AI protections that Azure offers.  

We have committed to building Azure into an AI supercomputer for the world, serving as the foundation of our vision to democratize AI as a platform. Microsoft pushed the frontier of cloud supercomputing technology, announcing our first top-5 supercomputer in 2020, and subsequently constructing multiple AI supercomputing systems at massive scale.  

 We also fine-tuned our purpose-built, AI-optimized infrastructure capability in partnership with OpenAI to train and deploy OpenAI’s family of models for research advancement and developer production. This infrastructure is now available to all Azure customers.  

“When other people come to us, we can literally give them the same style of infrastructure that we used for OpenAI, because that’s now the standard way that we do it,” Boyd says.  

Microsoft’s Azure OpenAI Service provides businesses and developers with high-performance AI models, such as GPT-3.5, Codex and DALL∙E 2, at production scale with industry-leading uptime. This is the same production service we use to power AI models in our own products, including GitHub Copilot, Power Platform, and the recently announced Microsoft Designer and AI-powered search in Bing and Edge.

We continue to evolve our AI infrastructure based on feedback and insights from training and serving AI models at scale. Our teams work in lockstep with industry partners on the design of GPUs, networks and datacenters that are optimized for AI workloads.   

“Microsoft continues to be on the cutting edge, and customers get to take advantage of all the benefits of that,” says Boyd. “They’re getting the best training infrastructure, the best software, the best networking – all of these things combined give the best experience.”  

Our approach to using AI for social good

Microsoft believes AI can help people tackle some of society’s biggest global challenges. Our AI for Good initiative provides funding, technology and expertise to help individuals and nonprofits accelerate progress in fields, such as accessibility, digital literacy and equity, sustainability and climate change, human rights and resilience, health disparities, food insecurity, cybersecurity and others.

“There are problems where AI is uniquely positioned to help, where AI is not just another solution but is the only solution,” says Juan Lavista Ferres, Microsoft’s chief data scientist and the director of the AI for Good Lab.

One such project, for instance, involves mapping all the places where humans live in order to better understand natural disaster risks and guide preparedness efforts. Another includes mapping rooftops of buildings in India, noting the materials the structures are made of to help disaster teams prioritize those that are more likely to fail in certain types of disasters. Yet another is mapping all the renewable energy installations around the world to help show the impact of solar and wind farms.

“A person would spend 400 years looking at satellite imagery and understanding if humans live there or not, but AI can do this in an hour,” Lavista Ferres says. “These AI models make it easier for people to get all the information in one place, and from there the humans can make the necessary decisions.”

Microsoft’s AI for Good Lab is an applied research and data visualization laboratory that harnesses the power of big data and Azure’s cloud technology. Its team of data scientists works with strategic partners and experts from academia, nonprofits and governments to not only help address critical global concerns but also to better measure the progress of efforts underway and to identify gaps where aid might be helpful.

The projects are as varied as the problems and combine Microsoft’s efforts with those of our partners.

For example, the AI for Health program invests in research led by institutions such as Fred Hutchinson Cancer Research Center, IRIS, the Novartis Foundation and Seattle Children’s Research Institute. These efforts are making healthcare more affordable, especially in places where doctors are few and medical needs are great.

Though we see great potential for AI’s assistance to help solve some of the world’s most pressing issues, we also recognize the limitations of the technology. It’s important to work through moral and ethical questions for every project, Lavista Ferres says, especially to make sure the models are getting enough data from enough places and aren’t leaving anyone out — for example, making sure all skin colors are represented for an app that detects skin cancer.

“When you train an AI model, it will be able to generalize from the data you have, but not from the data you don’t have, so any bias can generate problems,” he says. “Having humans in the loop is the key element. We are looking for solutions that can benefit everyone, not just particular cohorts of the population.”