Artificial intelligence innovation has become part of our everyday lives—retailers use it to tailor the product recommendations they make; biotech companies hope it can create customized medicine. But its shortcomings, born of human biases, are becoming apparent as well.
Take, for example, facial recognition technologies that work best on white people and make the most errors on black people. “Like a human, AI’s learning is only as good as the data it is fed,” said Manoj Saxena, former head of IBM Watson in Austin, TX, and a prominent investor in AI startups.
A new organization called Ada-AI, which is based in London, is focused on making sure the data used to build these technologies come from diverse sources. Specifically, the group seeks to make sure that non-Western, non-white parts of the world are included in influencing AI innovations. “We want to be thinking about who are the next generation of tech leaders,” said Madeline McSherry, Ada-AI’s policy lead. “A lot of them will be in Silicon Valley, but how can we encourage them in other areas?”
The Ada-AI effort has attracted to its board a number of prominent leaders in AI representing organizations such as NASA’s Jet Propulsion Laboratory, XPrize, and the World Economic Forum. “This is not about top-down, ‘we’ll first solve the problems in New York and Palo Alto and then export the solutions everywhere else,’” Amir Banifatemi, the AI lead for the XPrize Foundation, which runs a $5 million AI contest with IBM Watson, said in an interview. “The problems have to be solved as close as possible to the sources.”
The first project Ada-AI is tackling is setting up a high school specializing in science, technology, engineering, arts, and math (STEAM) in Herat, a city in the northwestern part of Afghanistan. The group is crowdfunding the effort and so far has raised just under $2,000 of its $500,000 total goal. (Ada-AI is also looking to raise funds from corporations and foundations for the
