I'm sure you've realized by now, if you hadn't before taking these courses, that machine learning has great potential to change society, and its potential to change society in a negative way. To increase inequality, oppression, exploitation, poverty, and desperation is just as great as its potential to change it in a positive way, to increase opportunity, justice, happiness, and growth. If anything, its potential for negative consequences is higher because often all it takes is haste and thoughtlessness for people to be harmed. As developers and machine learning practitioners, we need to put in the time and effort to think about how our Machine Learning projects can benefit society, how we can ensure that they're inclusive, and how we might prevent or mitigate any harm they might cause. In other words, to make AI for good and for all. I hope you realize that ethics and AI isn't just debating the trolley problem on social media, it's nuanced and contextual, and debate on esoteric thought experiments can take time, attention, and funding away from issues right here and now. Issues that can and do grow with little public attention or debate. Things like how targeted pervasive advertising is allowed to be. To what extent we use facial recognition, and whether it's acceptable to have an AI making decisions about resource allocation without human input. These are all pressing issues that have a practical, ethical dimension that impacts all of us. To help you in the process of understanding your own ethical stance, take a look at some of the AI ethics, frameworks, guidelines, and declarations, the different put out by different government and non-government organizations. We have a link to a comprehensive list of all those currently available in the class resources. Most of the current guidelines and frameworks are carefully and thoughtfully done. But our go-to, maybe because we're Canadian, is the Montreal Declaration. The Montreal Declaration was started as a broad collaborative process by the Universite de Montreal on November 3rd, 2017. The declaration is currently posted online available for anyone to read and sign. This isn't the final form of the declaration however. There never will be a final form because as AI and society constantly change, so must the ethical guidelines for the development and use of AI grow and adapt. Some of the principles of the declaration are things we've already mentioned in this specialization. Things like respecting the privacy of people using the system and the people whose data you're using, or how AI should encourage diversity and not restrict lifestyle choices, behavior, or opinions, or systematically limit opportunities for specific groups of people. But there's also a number of principles that we haven't addressed directly in this course or specialization as of yet that are just as important. Things like taking steps to make sure that your system can't be misused to cause harm, even if you yourself have no intention of using it that way. The principle that if your system is used to harm someone, you are ultimately responsible. Many AI researchers and developers have agreed they'll not work on military applications or autonomous weapons. Much like many physicists have agreed they'll not work on the development of nuclear weapons. It's not because of any existential threat from the AI, but because of how such tools are used. For example, the more layers of abstraction between me, as the person taking an action, and the people who would be harmed, the easier it is for me to act unethically. It's only human. This is why people say nasty things on the Internet that they would never say to someone's face. It's also why it's easier to kill people when there's a drone as an intermediary. Remember, your system is out there in the real-world impacting it, and sure, nothing can violate your principles if you don't have any principals. But that doesn't free you or the world from the consequences. So know what you stand for and where your boundaries lie. Establish your personal principals and make sure your organization stands by them as well. Knowing what principles you work by will help keep you from falling into the haste and thoughtlessness that ends with people being harmed. I know this is a heavy topic. It can be uncomfortable and it's natural to avoid that discomfort. But this is important, you will affect others. Plenty of people with good intentions are remembered for the harm they caused. Please, don't let your name beyond that list too. Machine learning has unprecedented potential and power. Let's use it for the betterment of everyone, for good and for all.
