Tech issues explained: Responsible AI

Hello again! Welcome back to this series from the editorial team that brings you Microsoft On the Issues . We’re breaking down topics at the intersection of technology and public policy and talking about how they impact your life. This edition: Responsible AI.??

When 2023 is discussed, a decade from now, the emergence of AI is likely to be one of the defining themes. AI has quickly gone from a utility behind the scenes to being a tool used by millions of people every day. When new technology has the potential to shape how people experience and engage with the world, it’s critical that it is developed responsibly.??

“AI may well represent the most consequential technology advance of our lifetime ,” says Microsoft Vice Chair and President, Brad Smith in a blog from February 2023.?

It‘s a big prediction, but Brad Smith believes AI will change almost everything we do. “Like no technology before it, these AI advances augment humanity’s ability to think, reason, learn, and express ourselves. In effect, the industrial revolution is now coming to knowledge work. And knowledge work is fundamental to everything.”?

The fact that AI is suddenly headline news will not have escaped you. The potential of these generative AI platforms has captured the popular imagination – with optimists and pessimists trading opinions on where AI technology will take us.??

Microsoft is a key player in the development of AI platforms – the potential of AI to do good and to facilitate harm raises a fundamental question: How can we ensure we develop responsible AI???

What do we mean by responsible AI??

In short, it’s about putting in place strong governance and safeguards. At Microsoft, these are embodied in a set of ethical AI principles that put people first .?

微软 ’s Chief Responsible AI Officer, Natasha Crampton , works closely with the engineers and researchers building Microsoft’s AI tools. Her job is to help ensure those principles are embedded right from the start of any AI development projects, building responsible AI by design. ?

Speaking in an episode of Microsoft’s tech and policy YouTube series TechTalk, Crampton says fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability come together to form a North Star for every AI project. ?

“AI systems are built by people, and they’re also used by people,” she says. “One area that is very important for us is human AI interaction. How can we make sure that we get the very best out of humans and machines working together? That’s Microsoft’s vision for how we embrace AI. It’s all about how we can amplify human potential. We need to figure out the optimum ways of combining the best of humans and the best of machines.”?

Security and privacy are essential building blocks for responsible AI. These must be protected if people are to place their trust in #AI platforms. The technology relies on vast amounts of data to train the systems, and it’s vital this data is kept secure. ??

Transparency and accountability are critical to developing responsible AI systems. It’s also important to be clear about the limitations of AI. People planning to use AI need to be able to understand how the systems behave and operate. ??

What is the potential of AI??

Human science can reap some of AI’s most powerful potential. It is already empowering medical scientists in their work to develop breakthrough medicines for diseases that are currently untreatable . At Novartis, AI is enabling researchers to speed up the development of new treatments by analyzing vast pools of data for patterns and trends that would take humans much longer to detect, if they spotted them at all.?

AI is also helping to create a more inclusive and accessible world . Scientists in Cambridge, U.K., have developed a system that enables blind and partially sighted people to locate other people in a room, facilitating more natural and rewarding conversations. Microsoft’s Project Florence initiative is working to replicate 540 million years of human evolution to develop a computer visual intelligence system with human-standard capabilities. This vision model will generate alt text to enable visually impaired people to engage with images on Reddit.?

As people reap the positive benefits AI will bring, there is also no doubt it will be used by people, communities, and nation states with bad intentions . For those who seek to spread disinformation, undermine democracy, and commit cyberattacks against individuals and other countries, AI is a potentially powerful weapon. Everyone from developers to regulators and law enforcement will need to work together to limit negative use of AI.??

Key goals in the development of responsible AI

Microsoft believes the development of responsible AI requires a broad dialogue that includes all stakeholders. This dialogue must focus on the positives and the negatives. Brad Smith has identified three key goals for advancing AI . The goals state:??

What policy is needed???

Throughout history, significant new technologies have demanded new laws and regulations to maximize technology’s potential while minimizing risks.??

Natasha Crampton believes regulators must continue to focus their attention on mitigating the potential downsides of AI, while allowing society as a while to ?reap the full benefits of the technology. ”I think that ongoing focus on high-risk use cases is really important,” she says. ?

In addition to legislation being discussed in Europe, the U.S. government’s National Institute of Standards and Technology (NIST) is consulting widely as it develops a framework for trustworthy and responsible AI .

Inward governance?

Developers of AI platforms will also need robust policies to regulate themselves. Microsoft is developing AI technology within a set of guardrails detailed in its Responsible AI Standard. In the interests of transparency, Microsoft has published this standard, along with a sample impact assessment, and sample transparency notes. Natasha Crampton gets into the details of the parameters and goals of the AI Standards in this blog post .

What you can do?

Read this:??

Take a look at Microsoft’s tools and guidelines that will help you create and operate responsible AI solutions, and continually assess their performance.??

Watch this:??

A discussion panel at the World Economic Forum 2023 in Davos, Switzerland, focusing on the role of investors in shaping the development of responsible AI.??

Do this:??

Natasha Crampton has shared three top tips to get started with ethical AI integration. These are:?

1. Think deeply about your use case.?
2. Assemble a diverse team to help test your product.?
3. Be committed to ongoing learning and improvement.?