Artificial intelligence is beside quantum computing one of the most exiting topics we actually discuss. For nearly all industries AI is promising massive changes, we looking forward to see it help us with predictive maintenance, simply support us in our daily life with knowing us and our dreams and wishes. Make our premises as secure as possible while processing sensor signals from various sources and take the right conclusions in real time.

Beside the technical promises AI’s are spooky. They can decide and act against human in autonomous drivers scenarios. Or even more famous the scene from the movie War-Games (1983) when the AI decided to destroy the world cause that is the only way to 100% win a tic tac toe game (Fiction only, its not to complex and can be won easily with the minimax algorithm).

Such scenarios made many people starting to think about methods to control AI’s and ensure that their decisions are not only correct but fit into our society narrative. An autonomous car should always kill an old man instead of a child perfectly clear situation and decision for us. A correct calculation can maybe get to a different result that is more correct but will never be accepted by a man.

Even bigger challenge will be a case where a AI take actions to ensure survive or distribute it self. Who will be able to notice that and is able to stop an endless intelligent AI that can be light fast in adjust strategies for that challenges.

I’m not sure if thinks like the Asimow Law somehow “hard coded” in the AI core processes will be able to stop such a synthetic organism. From my point of view such a highly intelligent “form of life” will be able to get over that hindering code to archive there objectives somehow. I can think of cloning relevant portions of the own code to an uncontrolled environment. The only challenge in that scenario will be a slow enough process to hide it from man before a critical portion of the data are “free”.

From my point of view the fact that a AI system will decide correct based on the available data instead of act as demanded by society can be the key observation. What if we connect all AI’s and let them create a society as we do it with children. We learn our children how to act in communities. We tell them the do’s and dont’s and basically teach them core values of our society. I see a chance to influence AI’s the same way. If we connect them to each other they can further interact and create there own narrative baseline hopefully based on the core values we touched them. At a certain point in time with a community of AI’s big enough to allow social interaction between individuals like we have, they should start to control them self.