The "WHEN Law" from The Three Laws Of AI
"Your wish is my command 'O Master..." the Genie said, when Aladdin rubbed the magic lamp. If we consider the Genie as an AI system then Aladdin is of course us, exploiting the might of AI. But our magical journey with AI, just like any good fairy tale, has to have limitations on WHEN we can unleash this magic. I had earlier proposed the "Three Laws of AI" that should be used to govern an AI ecosystem to helping us regulate and control its operation. If you are wondering about the topic in discussion you can refer to my earlier post on "The Three Laws of AI" to get some context on the link below.
The Second Law of AI that I also call the "WHEN law" is -
领英推荐
An AI system should do a task only when it gets an explicit approval from a human moderator to execute the task
The Human Moderator - A human moderator is someone human (duh!) who executes a task. In fairytale lingo, this is someone "Who rubs the Magic Lamp". Any man made system is always driven by a Human Motive to perform a task. Ever since we started inventing processes to execute a task, we have always needed someone human to execute the task and to evaluate its results. But this is no longer necessary! With increasing capabilities in AI, autonomous executions of tasks is now possible. The opportunities for automating are countless and its benefits, immense! But just because we can make systems autonomous doesn't mean we should. The risks of empowering an AI powered system to take care of everything for us far out weighs the benefits. Just imagine a case where your smart watch that monitors your vitals notices you have been under a lot of stress lately. So in order to help you relive your stress, it co-ordinates with your travel app to look up your favourite holiday destination, and then goes ahead to check your calendar to pick the best days and finally teams up with your bank to book you the perfect holiday, without ever consulting with you or seeking your approval. Believe me all of this is possible today, and since the intent is noble there is nothing stopping this from happening. But would you approve of such a behaviour? While my example here is benign there could be a lot many more use cases where the outcome could be a lot more worse.
Uncle Ben said "With great power, comes great responsibility". Gauging from how things are shaping up, it's evident that we are endowing our AI systems with Great Power, but what about Responsibility? Is it even possible to codify Responsibility and make it a part of the AI's architecture? Till we get a convincing answer for that question, I would like to lean upon the second law to assure me there is always someone Human with a moral responsibility governing WHEN the magic lamp is rubbed!
As everybody else I would love to hear your thoughts through the comments and feel you love through the like and share buttons. So please don't be shy!
#AI #artificial #intelligence #rules #governance #mind&machine #laws #robotics