One of the “godfathers” of artificial intelligence is calling on the City of Toronto to sign a declaration for the responsible use of AI amidst concerns over data privacy, police surveillance and Sidewalk Labs’ waterfront smart neighbourhood.
With the ability of sensors to harvest massive amounts of data, and the increasing power of artificial intelligence to derive insights and predictions from that data, city services like public transportation, waste collection and snow removal can all be optimized, says Yoshua Bengio, a computer scientist at the Université de Montréal.
“This would be great. The concern is that measuring all that data has to be done in a way that is consistent with issues of privacy, control and democracy. That’s why we want to put in these principles, and make sure that civil servants understand them and don’t deploy things that go against these principles.”
Bengio was one of the computer scientists whose research on a type of machine learning called neural networks led to unprecedented breakthroughs in the ability of computers to carry out tasks like image and speech recognition. Those advances underpin modern AI, from virtual assistants that recognize verbal commands to social networks that automatically label faces. Last month, Bengio shared the 2018 Turing Award, sometimes called the Nobel Prize of computer science, with two other neural net pioneers, Geoffrey Hinton of Google, the Vector Institute and the University of Toronto, and Yann LeCun of Facebook and New York University.
Last December Bengio helped organize the Montréal Declaration, an ethical framework for the development and deployment of artificial intelligence. The declaration identifies a set of values and priorities to guide AI, from the protection of privacy and democracy to the maintenance of social and cultural diversity. The City of Montreal is one of 41 organizational signatories; 1,178 individuals have also added their names. Bengio believes the City of Toronto should sign on, too.
A spokesperson for Toronto Mayor John Tory said “the mayor has asked city staff to review the declaration and report back on the implications of signing it.”
“As Toronto continues to grow its innovation and technology sector, it is critical for the City to work with leaders in the industry to learn and adopt best practices,” Don Peat added in an email.
Bengio, who was in Toronto for a panel discussion on ethical AI, believes Canada should be a natural leader in this area because of its values.
“We want Canada to be united behind the goal of an ethical AI,” says Bengio. “Every sign we send to the world that we stand strongly behind these values is going to help, both internally and externally: Internally, because we are telling our citizens, our companies, that we take these concerns seriously, and externally, because we have a chance to influence in which direction the world is going with the use of AI.”
Artificial intelligence, if deployed with care, can be a major social boon, Bengio says. Computers have already demonstrated the ability to quickly and accurately analyze medical images, he notes, and his lab is using neural networks to tackle climate change and create novel molecules for drug development.
But the risks are great too, he says. He worries about the development of autonomous weapons, like a drone that is authorized to kill if it recognizes the right face. Profit-driven companies can use data derived from social networks to influence your behaviour, or worse, your vote. He believes the social safety net needs to be strengthened for workers who will be made redundant by algorithms, and that too much power is being concentrated on the campuses of a few technology companies. (He dismisses fears about the oft-cited “singularity,” in which a supercomputer becomes more intelligent than humans and destroys us, noting that machine-learning algorithms still lag far behind the general intelligence of any 2-year-old.)
In the context of cities, Bengio recently co-signed a letter to Amazon asking the company to stop selling a facial recognition software to law enforcement. The software had much higher error rates when identifying darker-skinned women compared to lighter-skinned men, raising the risk of police misidentifying racialized women as suspects.
In Toronto specifically, police recently acquired cellphone surveillance technology that captures data on hundreds or thousands of bystanders’ mobile devices in addition to that of criminal suspects. Bengio believes we need to be “extra careful” to guard against privacy violations with the data police collect, and that “the decisions to use new technology should go through public scrutiny.” (Toronto police have cited a non-disclosure agreement in declining to answer questions about the technology.)
Bengio also says he understands critics’ concerns about Sidewalk Toronto, the parcel of waterfront property that Google parent company Alphabet’s Sidewalk Labs is designing as a smart neighbourhood. Concerns have been raised about control of the data collected in that project.
“I understand the concerns, and I think we should be careful and make sure that it doesn’t happen behind closed doors — that the city or a representative of the people have a clear view of what’s going on, some transparency.”
A Sidewalk Labs spokesperson said they have reviewed the AI declaration and agree with its principles.
This article originally appeared in The Star