“See Jane run. Run Jane run.” Reading a book. That was the culmination of a whole year of kindergarten and occurred at the start of grade one. Kindergarten was spent on learning to recognize the letters of the alphabet, drawing them and sounding them out. It never occurred to me to put the letters together to make words. That was the brilliant idea of my grade one teacher. That day was like magic.
I was hooked, I could not stop. My education went from real life experiments of eating dirt, riding a bicycle and wandering the neighbourhood to emptying out our school library and waiting for the next bookmobile to stop by the school so I could borrow some new books. The cat was out of the bag, never to return. Which makes me wonder, when were cats ever in bags to start with, I hope the expression was created before plastics.
That human processing is slower than that of the technology we create is an understatement. With ChatGPT out of the bag, we wonder both how AI can help us perform our job (or studies) better and when it will replace us. And then, of course, no longer in the realm of Arnold Schwarzenegger movies such as The Terminator, when will it destroy us, if we are not careful. Arnold himself now admits he was the product of drugs we discovered to make a “better” human, making him the perfect choice for a cyborg, the next level of “human”.
AI is the product of human creation and, unfortunately, it also comes with our many biases. Garbage in, garbage out, so to speak. Humans have conscious and unconscious biases drilled into our biological systems. We find all sorts of patterns. AI results from combining large data sets with machine learning and analytics. Feed it photos of endless cats in every position and iteration and it will learn to identify cats. It looks for those patterns.
People might assume that smart people working in AI would avoid or control for biases and provide us better results than we ourselves achieve, but they might not even realize their biases, or the data in existence or accessible might have inherent biases. If the data sets are based on historic information that excluded certain races or included mostly one gender then it may feed you results, for instance, that favours doctors being white males, perpetuating stereotypes. You might not be granted a loan or interview as your data is not selected based on your postal code reflecting a certain neighbourhood or your choice to stay home to raise your children creating a gap in your credentials.
For recommender systems, I try to look at news on both sides of the spectrum (as if there are only two, but only so many hours in a day) so that I might get both sides of the news fed to me and guess a happy middle. To keep you clicking, they try to give you what you want and you might want to think it is therefore the truth. Don’t.
Maybe like cigarettes, these products should come with warnings. For instance, when you select your self-driving vehicle, its ethics should be disclosed – when faced with two deadly decisions, which one will it make – save the pedestrian or save your passengers? This brave new world is upon us but not only the scientists creating AI should have a say.
Donna Purcell, K.C., (aka Lady Justice) is a Central Alberta lawyer and Chief Innovation Officer with Donna Purcell QC Law. If you have legal questions, contact dpurcell@dpqclaw.com.