While many life changing technologies such as the iPhone are created in windowless rooms away from prying eyes, firms developing artificial intelligence may just be seeking to share their progress. In particular, their failures.
The Partnership on AI, a consortium organizations includes firms such as Apple, Facebook, BBC, and ACLU, is currently building a database that collects incidents of misbehaving A.I. In doing so, the Partnership hopes companies developing the technology can better learn from each others mistakes, said Terah Lyons, executive director of the Partnership during the Coins2Day Global Tech Forum in Guangzhou on Thursday.
“One of the products we’re building is a database of incidents, basically failures of A.I. To operate correctly in certain environments,” said Lyons, previously policy advisor the U.S. Chief Technology Officer during the Obama era.”Our hope really is that will become a research resource to the community and publicly available.”
While that database is currently being put together by the airline industry, the entire group of 80 plus participants could in the future use the database to tackle not only technical issues, but also ethical questions about how to keep human biases from affecting the algorithms.
Lyons is hoping to release the database to the public in early 2019.
Not only is the partnership trying to open doors across industries when it comes to mistakes in AI, it’s also hoping to develop a broader base of partners. Most recently, the consortium added its first China-based firm: Baidu.
“We’re in the business of productive disagreement, so I think airing better, more nuanced understandings of cultural, jurisdictional, political policy differences in the environments in which the different organizations that we work with operate and the products and the values that they develop themselves is really important in what I see as a very global and human project about getting this right,” she said.