AI the NOT Rosy Side

A university student integrated an AI brain in a coffee machine with one objective, “be always ready to serve coffee”. This AI brain realized that the “switch-off” button is conflicting with his objective, so, it disabled it. Then it anticipated that the humans walking near the coffee machine could also “conflicting with his objective”, so, it run reinforcement learning algorithm … to disabled them!

“If” such AI model did any harm to the humans, it is NOT due hard feeling to them, and NOT due any sort of self-awareness, or the singularity concept … it simply doing what it was designed to do. “Bad design” in IT application could lead to system crash, or error message, but “bad design” in an AI model that was integrated in self-driving car, smart house, or military equipment could lead to catastrophic results!

Professor Stuart Russell is one of the first to highlight this issue; he stated that AI is very good “objective-achieving machine”. So, if a human designed -accidently- bad objective, then AI will work perfect achieving that objective. This why there MUST be general roles controlling such AI super power.

Lately, Professor Stuart conducted a series of four lecture for the BBC Reith Lectures: “Living with Artificial Intelligence”, covering AI in warfare, economy and AI a future for humans. During these lectures he walk the audience through all these challenges. I really recommend you to listen to them, very informative & insightful lectures.

Professor Stuart state that the challenge of AI harming human will not be solved by AI experts, but actually by philosophers, poets and writers … until then, if you know that a coffee machine has embedded AI technologies … ensure to take your cup with extreme cautious!