Training robots, as well as being fun, should include input of people from all genders and ethnicities, to avoid bias
AI is only as good as the algorithms that feed it. As a result, the technology can be biased, depending on how it has been trained. In certain circumstances this bias can have drastic results.
“A biased AI controlling an automatic door might open for white men, but not for black women,” explains Samira Samadi, leader of the Fairness in Machine Learning group at the Max Planck Institute for Intelligent Systems in Tübingen. “Errors like these have an effect on people’s lives.”
Samadi’s institute is part of Cyber Valley in Baden-Württemberg, in the south-west of Germany, an area that has become a beacon for AI research and development. And one of the big aims for Samadi, and others like her, is to make AI more equitable by eliminating the sources of bias so that the algorithms can work for everyone.
Collaboration and partnership across the valley
Nestled in the mountains bordering France and Switzerland, Cyber Valley was founded in 2016 by partners from government, science and industry (see ‘Partnering for AI research’). Just five years later, the region already contributes more research to machine learning and computer vision conferences than anywhere else in Europe.
“There are lots of people conducting good research in Cyber Valley,” says Wieland Brendel, AI group leader in the department of neurology, University of Tübingen. “There is also a growing community of start-ups, so there is research outside academia,” he adds, who is also co-founder of one such start-up called Layer7 AI.
Such close ties between science and business have helped propel Baden-Württemberg to the top of the EC’s European innovation scoreboard for 2021.
But, Brendel explains, AI isn’t just a matter for those who create the technology, it also connects with society. Researchers should consider the effect of technology on a wider scale. In March 2021, Cyber Valley launched ‘AI and Society: Exploring Possible Futures’, a public engagement initiative to make AI more accessible by giving members of the public the chance to meet AI researchers.
This initiative complements Cyber Valley’s Public Advisory Board, which reviews the ethical and social implications of projects funded by the Cyber Valley Research fund. “We have a responsibility to be outside our research bubble,” says Brendel.
Samira Samadi, leader of the Fairness in Machine Learning group at the Max Planck Institute for Intelligent Systems in Tübingen.
Building a future in Baden-Württemberg
As well as connecting research, Baden-Württemberg also connects people as it welcomes scientists from all over the world. Samadi was born in Iran, and moved to Tübingen in September 2020, following her PhD at Georgia Tech in the United States. Here she has found a new home.
“Tübingen is a good place to be a young computer scientist. It’s growing fast, and the funding is there for AI,” says Samadi. “It is one of the best places to be for research.” Baden-Württemberg invests heavily to maintain this position. In 2017, it contributed 5.6% of GDP to research and development, the highest of all 78 regions in the EU.
But there is more to Cyber Valley than just its professional appeal. “The community is welcoming and friendly,” adds Brendel. “It’s a good place to have a balance of family life, research and entrepreneurship.”
