Microsoft making AI human

Microsoft is focusing research and development on ensuring AI can have a positive impact on people and society

Tags: Artifical intelligenceMicrosoft CorporationUnited Kingdom
  • E-Mail
Microsoft making AI human AI has great potential, says Shum, there also needs to be understanding of how it will relate to humans.
More pics ›
By  Mark Sutton Published  July 24, 2017

There have always been concerns about Artificial Intelligence, and as the technology has moved from development into real world deployment, so it is becoming clearer that there is potential for AI to be a negative, as well as a positive force. Now Microsoft, one of the leaders in AI development, is looking to ensure that with the wider adoption of AI, there is also a proper consideration of how the technology will work with people and society.

At an event in July, Microsoft announced a number of new initiatives around AI, which are intended to boost artificial intelligence research and emphasise its role in developing people and society.

Among the new initiatives for Microsoft is the formation of Microsoft Research AI (MSR AI), a team focused on diverse areas of AI development, and the AI for Earth Program which will make AI technology and tools available for free to organisations focused on sustainability and environment projects.

Harry Shum, executive VP of Microsoft's Artificial Intelligence and Research Group, said at the event that Microsoft is aiming to focus on positive usages of AI, that will ensure the opportunities of AI are matched with proper consideration of how people will use and interact with AI, and how it can impact on society.

"AI has so much potential to improve our lives, but with all the great opportunities, come many tough problems, such as jobs and security. It all comes down to one thing for us at Microsoft when we think about AI, it is about people. It is about amplifying human ingenuity through intelligent technology. This is our guide for everything we do as a company," Shum said.

The next big step for AI is to move from AI that can perform very specific tasks, to developing more general capabilities and intelligence, to be more human-like in the way that AI can learn and interact. This will require AI that can understand and interact with humans in two-way dialogue, but this is a very difficult thing to do, he noted.

With this new frontier of AI come the need for universal principles that guide thinking in design and implementation of AI, Shum added.

"Our goal is very simple - we want to avoid a situation like HAL from 2001: A Space Odyssey. It was this belief, that led [Microsoft CEO] Satya Nadella to outline a series of principles to influence an empathic and ethical approach to AI design industry-wide, since then we have simplified with the acronym of FATE - the fairness, accountability, transparency and ethics of AI design. We recognised we could not do this alone, so we helped to found the Partnership on AI"

The Partnership on Artificial Intelligence to Benefit People and Society (Partnership on AI for short), is a non-profit industry body launched in September last year, to address policy and ethics around AI. Member companies including Microsoft, Amazon, Google, Facebook and IBM will work with scientific and academic communities to conduct research and develop best practices and open licences in areas such as ethics, fairness, and inclusivity; transparency, privacy, and interoperability; collaboration between people and AI systems; and the trustworthiness, reliability, and robustness of the technology.

The group also proposes to work with professional and scientific organizations, such as the Association for the Advancement of Artificial Intelligence (AAAI), as well as non-profit research groups including the Allen Institute for Artificial Intelligence (AI2).

Eric Horvitz, founding co-chair of the Partnership on AI, Technical Fellow will head up the new MSR AI as its managing director. The MSR AI will bring together around 100 engineers and researchers, the "top talent" in AI, Horvitz said, to apply their diverse expertise to a range of critical research areas in AI.

When the term AI was first used in 1955, it was related to key ‘human' attributes of intelligence - perception, learning, reasoning and natural language - and the team will focus on these areas, the intersections between them, and breaking down silos of AI research. While there have been advancements in AI since 1955, it is really in the past eight years that research has made great progress, due in part to areas like enhanced computational power and the ability to work with large data sets.

This progress will result in AI having a transformative influence on many key sectors, including, health, agriculture, education, transport, Horvitz said, and MSR AI will look not only at developing AI capabilities, but also at how human-AI collaboration can be improved, and how AI systems that have been developed in a closed laboratory will operate in real world, open situations.

MSR AI will focus on creating insight into the area of AI, people and society, Horvitz explained: "There are several critical directions, as AI is going to touch humanity in myriad ways, where we have to bring in psychologists, experts in human factors like economics and safety critical systems engineering; sociologists and social scientists, to understand how to do this well, how to address possible adverse outcomes, and rough edges. How do we build robust systems that work in the open world, even when they were trained in narrow laboratory worlds?"

The work in AI, people and society will look at principals including trustworthiness and safety; fairness and accuracy; transparency and explanation and exploring human-AI relations. There are many different potential areas that need to be addressed, Horvitz said, such as ensuring privacy of data, promoting understanding of how data will be used, or understanding that data can be biased or skewed, and might therefore create issues if that data is used in AI decision making. MSR AI will aim to consider all of these issues, and also to collaborate openly with others in the field.

These wider considerations for AI development will also be addressed at the highest level within Microsoft, Horvitz added.

"As part of our work to think through AI, people and society, we have created what we call the Aether board at Microsoft, which stands for ‘AI and Ethics in Engineering and Research'. This is a senior level board that reports to our CEO. The idea is that it will continue to consider rising issues in the realm of AI, people and society, bring opportunities and challenges to the attentions of the company's leadership and make recommendations and help formulate policies and technologies.

"I am extremely excited and optimistic about where things are going, my optimism is coupled with a sense that we need to be responsible, at the top of what do at MSR AI is to pursue the principles of the science of AI."

Add a Comment

Your display name This field is mandatory

Your e-mail address This field is mandatory (Your e-mail address won't be published)

Security code