With the recent breakthroughs in deep learning and computer vision, we are beginning to see Artificial Intelligence being used in applications that have once existed only in science-fiction settings. Our AI computer vision algorithm has been deployed in seniors homes to prevent seniors living with dementia from going missing; our advanced computer vision algorithm is now integrated with infrared-sensing arrays to conduct high-throughput fever detection for the purpose of preventing the spread of COVID-19.
Understanding of the legitimate concerns for AI is the key
Yet, we, collectively as a society, need to be mindful about how advanced technologies are applied appropriately. Specifically, we must prevent the misuse and abuse of advanced technologies like AI.
AIH and its partners have been advocating the principle of precision regulation. This approach involves carefully targeting legislations to address legitimate concerns on AI while encouraging technological advancements and building a culture of accountability within the AI community. We strongly believe this approach would assist lawmakers address the privacy and civil rights concerns that have been raised on AI.
In the context of facial recognition, it is imperative for both technologists and policy makers to understand the significance of our faces to our personal identity as individual human beings. We must see the perspectives of those who have reluctance in accepting facial recognition technology in certain applications, for example, identity authentication. However, an outright ban on the technology should not be the immediate go-to solution to address these concerns. Such regulatory approaches could result in cutting many critical benefits, including those that are life-saving, facial recognition could bring to society. Take the discovery of nuclear fission as an example: the tragedies of Nagasaki and Hiroshima should not prevent nuclear reactors that are producing the life-saving nuclear medicines from curing cancer patients with aggressive malignant tumours. Facial recognition, too, should be approached with a precision-regulation framework.
Be informed on the technology before legislating
First step in a precision regulation approach is to start with an educated understanding that not all applications involving facial recognition can be lumped together under the same umbrella. Specifically, the applications of facial recognition algorithms can be categorized into three groups:
- Face detection: this involves recognizing that an object in an image is a human face. However, it does not identify who the face belongs to. Face detections have been used for applications that involve in human pedestrian traffic analysis and people counting, which could be useful in ensuring fire safety compliance in a facility.
- Facial authentication: this refers to a system using facial features as a means of identification. Most new smartphones have adopted this technology to help users unlock their phones. From the engineering perspective, this technology involves a 1:1 facial feature similarity comparison, which does not require heavy computation capabilities.
- Face matching — this refers to matching one or many faces to a group of face images stored in a database (1:N or N:N comparisons). Face matching has been used in AIH’s senior wandering prevention program — in which we use to prevent specific seniors (i.e. those living with dementia) from going missing at the seniors homes. Additionally, this application could be used as an effective victim identification tool in human trafficking and child exploitation investigation cases.
Each of these applications require a precision-based approach in ensuring the technology is applied appropriately, without misuse and abuse. Instead of putting out an indiscriminate outright ban for the technology, policymakers ought to consider the application and end-users. A broad regulatory framework that fail to differentiate between these applications will lead to unintended consequences that may not be the legislative intent of the lawmakers. An outright ban on the use of facial recognition by a municipal government could prevent victims of human trafficking and child exploitation from being identified and saved in a timely manner. In combating the opioid crisis, such ban would prevent emergency first responders to rapidly identifying unconscious or non-verbal patients and give them the proper life-saving treatments in time.
However, we do believe that certain applications of facial recognition technologies should be strictly prohibited. The practice of scraping social media photos without the consent of the users and running facial recognition on those images for private corporations and government agencies is a clear violation of privacy law and civil liberty. We do believe that companies involved in such applications should be brought to justice and be held accountable for their irresponsible actions. The basic principles of the rule of law must be upheld.
Lawmakers should not wait another minute to take actions addressing public concerns over facial recognition. However, enacting a law with the effect of outright-banning the technology is neither constructive nor practical. Used appropriately, advanced technologies could save time, save money, or save lives for the communities across Canada. Facial recognition is no different. Government, technology companies and community stakeholders must work closely on delivering practical solutions to address the real-world problems. This is a conversation AIH and its partners