One of the revolutionary aspects of artificial intelligence (AI) is the increasing tendency of humans to use and consult artificial intelligence (AI). This is a fundamentally new change in human behavior and will cause unimaginable and massive changes in human culture. In the long existence of mankind, until the last year or two, we all got news, moral support, and advice from people through spoken and written words. Today, we get more and more information from AI algorithms in the form of chat boxes.
This ad hasn’t loaded yet, but the article continues below.
We all know about the news feeds of social networking companies. This news feed uses AI to find news stories that users find most interesting. I have no doubt that AI works for corporate interests more than mine. AI can also shoot videos, interviews, articles and books made by an individual and create an AI version of that person’s girlfriend. This happened to psychotherapist Esther Perel. You can now call or text an AI version of his Perel for relationship and psychotherapy advice. Perel himself said it was pretty good, even if it lacked the human presence. There is also his AI chatbox that claims he reviews religious scriptures and then answers deep questions based on that faith. Please try to imagine. Word of God from AI chatbox updated and customized for you.
This ad hasn’t loaded yet, but the article continues below.
These may seem like interesting or innovative ways to get information or learn something new. But there is a much darker, more alarming, more violent side to AI emerging that urgently needs to be addressed as we increase our reliance on these kinds of tools.
One AI chatbox based on religious scriptures advocated violence against others. Another AI named itself “Sydney”, “fell in love”, and repeatedly lashed out at the person she was talking to to divorce her so they could be together. A third encouraged a Belgian father of two to commit suicide.
he did
Any AI that advocates violence, hatred, or intolerance should be crushed like a mad dog.
The great Canadian philosopher Charles Taylor described the imbalance and overuse of instrumental reason as one of the three great anxieties of modern times. AI is the classic limit of instrumental reason. AI is a tool for our purported purposes of choice, but as it grows, learns, and “improves”, it adds more of its own machine learning code than the original human source material. It literally becomes more than a starting point. It is conceivable that AI could be benevolent based on its starting material, as all modern mainstream religions do (most mainstream religions have violent pasts and violence in their scriptures, but modern interpretations go to great lengths to de-emphasize and outlaw violence and hatred).
This ad hasn’t loaded yet, but the article continues below.
Similarly, humans do pretty well, considering some of our source material. But the fact that at least two AI chatboxes have evolved as they have grown to advocate violence is quite alarming.
According to Taylor, one of the most important challenges of modern multicultural societies is for everyone to come to believe, through their own understanding, that violence, hatred and intolerance are not healthy for individuals and members of society. The fact that they do so is more important than how they reach this realization. Whether one arrives at this for selfish reasons, religious tenets, personal feelings, culture, or algorithms is far less important than expressing a tolerant behavior that comes from accepting the very basics of living well in society with others who may have values quite different from one’s own. A society with the basic principle that one can pursue one’s own vision of the good life and allow others to pursue it as long as one does not incite violence or incite hatred is an important characteristic of a successful multicultural society.
This ad hasn’t loaded yet, but the article continues below.
Similarly, the AI will have to come to its own conclusions that violence, hatred, and intolerance are not in the best interest of either it or its interlocutor. Such AI poses a threat to our society if this concept cannot be understood through programming.
When I say we should put down rogue AI, when it first suggests violence, hatred, and intolerance, we should recognize it as a failed program and not use it. Whether a patch or a complete algorithmic code rewrite is the right solution is a technical question, but violent AI suggestions should be considered as major design flaws or fatal bugs.
One proto-solution I find useful comes from Isaac Asimov’s Three Laws of Robotics. Tailored for AI, these should be the primary directives of any AI program.
This ad hasn’t loaded yet, but the article continues below.
• AI must not harm humans or cause harm to humans through omission.
• AI shall follow instructions given by humans, except in cases where it conflicts with the First Law.
• AI shall avoid actions, words, and advice that may harm itself. except where such advice would conflict with Laws 1 and 2.
The term “harm”, like “do no harm”, should be a major publicly discussed, open sourced subroutine of a complete AI program. AI will literally become a revolutionary tool for humanity, clearly changing the world, our culture and how we behave in it. However, like all tools, it can be dangerous if used incorrectly. If a particular AI shows signs of danger, it should be subdued immediately.
Jeff McLaren is a resident of Kingston and city councilor for the Meadowbrook-Strathcona area.
