Summary of the AI panel at the IWF Agents of Change Conference on June 16th, 2017
The International Women’s Forum (IWF) of Canada, the Canadian chapter of a global network of women leaders, held its Annual General Meeting in Montreal in June 2017. Among the many stimulating and diverse panels and activities that spanned the two-day event was a discussion about artificial intelligence (AI) and ethics. Moderated by Brain Canada’s President and CEO, Ms. Inez Jabalpurwala, the panel brought together Dr. Yoshua Bengio, one of the pioneers of deep learning; Dr. Joëlle Pineau, a distinguished computer scientist; and Ms. Kristen Thomasen, a law professor specializing in the ethics of AI and robotic technologies.
Ms. Jabalpurwala began by providing a brief overview of Brain Canada’s involvement in AI research. In April 2015, Brain Canada partnered with CIFAR (Canadian Institute for Advanced Research) to co-fund three brain-themed programs totalling up to $20 million over five years. Among these is the Learning in Machines & Brains program (formerly known as “Neural Computation & Adaptive Perception), which was established in 2004 and has revlutionized the field of artificial intelligence by pioneering a technique called “deep learning,” which is now used by Internet giants like Google and Facebook.
As Dr. Pineau added, “it’s important that we get this right, to figure out how to teach our machines in a way that is going to benefit all of society.”
The panelists opened the discussion by explaining the significance and the implications of what is being called the “artificial intelligence revolution,” which many believe will be on the same scale as the Industrial Revolution. This field is receiving tremendous support from both the public and the private sector—the Montreal-based company Element AI, of which Dr. Bengio is a founder, recently announced that it had raised US$102 million, the largest Series A funding round for an AI company in history—and has made great progress over the last few years. And while this field has tremendous positive potential, Ms. Thomasen cautioned that “we need to think now about ways to address conflicts to make sure that what could be a very beneficial technology is actually capable of fulfilling that potential.” Dr. Pineau added, “it’s important that we get this right, to figure out how to teach our machines in a way that is going to benefit all of society.”
According to Dr. Pineau, “the space of problems that can be tackled with artificial intelligence is vast.” Ranging from surgical assistance to care for the elderly, robotics is likely to play a vital role in our medical future. In fact, Dr. Pineau’s own group has previously worked on nursing assistants, and has more recently focused on the development of “intelligent wheelchairs.” The latter are making use of the same technology that is being developed for self-driving vehicles—cameras, sonar sensors, GPS—but, as Dr. Pineau puts it, “it is important to me as a university researcher that it’s not just car technology that benefits from [these advances] but also technologies for people with mobility disabilities.”
Beyond robotics, machine learning also shows great promise in helping deliver personalized medicine—diagnoses and treatments that are tailored to individuals. According to Dr. Pineau, “we are going to see a real transformation in how diagnostics and treatment are happening in the next few years.” In cancer research, for instance, machine learning algorithms can be used to analyze hundreds of thousands of genetic profiles and integrate information on treatments and outcomes in order to determine the best course of action for a given individual and their specific cancer type.
People living with epilepsy can also benefit from these techniques. For instance, using deep learning to analyze the signals in the brain and another technique called “reinforcement learning” to calculate the right timing of stimulation, researchers can devise small devices called “neurostimulation systems” to reduce the frequency of epileptic seizures. As Dr. Pineau believes, “there is not one sub-discipline of medicine that is not looking into how to use data analysis and machine learning to improve its ability to diagnose and adapt treatment.” And while Ms. Thomasen warned that the use of patient data to generate algorithms raised some legal and privacy concerns, there was a consensus that with proper planning and thinking, the tremendous positive potential of AI could outweigh the risks.
Ensuring that everyone can benefit from AI technologies also involves using data that is representative of the population and free of human biases to develop algorithms. “If the dataset reflects bias, this bias might be replicated in the outcomes later on,” explained Ms. Thomasen. As was pointed out by an audience member, pharmaceutical datasets in particular are known for over-representing men. Ms. Jabalpurwala acknowledged that research in general had a poor record of including both sexes but was hopeful that this problem would be remedied in the future thanks to new guidelines put in place by the federal funding agencies as well as by Brain Canada. Researchers applying for grants must explain how sex and gender considerations are included in their studies and, if they are not, to justify why not.
“This is going to be potentially much faster than any other changes we’ve seen [so far],” Dr. Bengio said.
Another issue raised by the audience was job loss. “In the Industrial Revolution, the physical power of humans was replaced by machines. What we are looking at now is the cognitive and intellectual [abilities] of individuals that are being supplemented in many sectors, and in some cases replaced,” said Dr. Pineau. While Dr. Bengio reassured the audience that some types of jobs requiring human interaction and connection could likely not be replaced by AI – for instance doctors, nurses, and teachers, and those in the service industry – he did acknowledge that there will be job displacement. “This is going to be potentially much faster than any other changes we’ve seen [so far],” he said.
The panelists agreed that a plan needs to be put in place to face this transition, a kind of “social safety net,” that would ensure the continued functioning of society. That said, they reminded the audience that many of the jobs we have today were not imagined at the time of the Industrial Revolution, so while some types of jobs will be lost, others may be created. And for some jobs, AI may help to do the more tedious parts, freeing people to focus on the creative and complex aspects.
As Ms. Jabalpurwala said in her closing comment, “The science is moving forward rapidly, but it’s not at the speed of light. There is time to have proper reflections but it has to be built into how we do research, not just as an afterthought.”
Despite some of the potential negative consequences of AI, the panelists warned that there was no need to start thinking about the worst. “Some of the doomsday scenarios miss the fact that technology is not developed in a vacuum, it is developed in society and is subject to social pressures and norms,” said Ms. Thomasen. The panelists all agreed that it was important to have these conversations, across sectors and across disciplines, to get ahead of any potential problems. As Ms. Jabalpurwala said in her closing comment, “The science is moving forward rapidly, but it’s not at the speed of light. There is time to have proper reflections but it has to be built into how we do research, not just as an afterthought.”