Diversity in AI Research

Artificial intelligence (AI) has entered everyday lives in many ways such as web search, social media, and virtual assistants. Along with entering daily lives of ordinary people, it has penetrated in almost every sector in the world. From BPO to manufacturing and from law to automobile, AI has made its mark in transforming processes and layoff of employees performing mundane tasks. The development of AI has resulted in ease and convenience in many tasks along with increase in efficiency. While the AI has been developing and progressing rapidly, there is a problem which did not come into much light: lack of diversity among AI researchers. This problem has been making AI biased inadvertently. There is a possibility that solutions to problems of people with different races are not considered. If there are black researchers in the AI research field, they would present problems faced by them and try to find solutions for them.

Timnit Gebru is endeavoring to counter lack of diversity in AI research field as a part of Microsoft’s Fairness, Accountability, Transparency, and Ethics in AI group program. Moreover, she cofounded the Black in AI event at the Neural Information Processing Systems (NIPS) conference in 2017. She pointed out that problems faced by black people will not come into light if there are no black researchers. As a result, AI will be biased to kinds of problems researchers think are significant, kinds of research studies they find important, and the direction of progress. When problems do not affect much, they are not given much of importance. Researchers may not know some of the problems because they have not interacted with people who have been facing them.

Ms. Gebru highlighted that there is a diversity crisis in AI research field. Along with discussions about law, ethics, and technical aspects, there is a need to have conversations about diversity in AI. This crisis needs to be treated as urgently as possible. She has been working on finding the way to encourage companies in providing more information to users or researchers. Companies need to determine if data sets they are using is biased and what their pitfalls are. She pointed out that AI research field is moving toward the mainstream, in almost every product there is. So, there must be a discussion about usage and standardization.

Timnit Gebru was on the steering committee for the first Fairness and Transparency conference held in February. She addressed the issues such as fairness, ethics, accountability, and transparency in the field of AI. Workshops have been organized at machine-learning-based conferences and natural-language-processing-based conferences. It is significant to have a stand-alone conference on aforementioned issues and people from various disciplines need to be talk about it. Researchers working in machine learning field will not solve this problem alone. Moreover, there are issues of transparency and updating the laws. If the unbiased approach needs to be adopted, they need to involve people who are facing problems due to bias.

There were very less black people working in the AI research field when Ms. Gebru counted at NIPS. She was informed that there are nearly 8,500 people working in the field. But she could see only six people at NIPS. AI has been impacting every part of society and moving into the mainstream. So, the focus on diversity must be given to ensure the problems from various ends are covered.