Today, the business value of AI consists of generating information from information the business doesn’t have: AI lets the business predict which solution, product or candidate is likely to be most successful. AI uses algorithms making sense out of unlimited data, taking over activities from simple to highly complex roles. If AI has access to data, it performs.
AI will be taking over more than simple or difficult parts of your job. Activities requiring uniquely human characteristics and years of experience are within its grasp. AI is becoming of age: solving many problems on its own and being self-aware. In our effort to create machines able to learn for themselves we not only gave them deep learning, computer vison and all kind of robotics, but also the capability to process natural language and learn through reinforcement. The future business value of artificial intelligence will not limit itself to removing a cost item, like the driver in a self-driving car or a pilot in a self-flying plane. It will come from self-governing, self-aware, autonomous intelligent machines. Mark Riedl and Brent Harrisson, researchers at the School of Interactive Computing at Georgia Institute of Technology, have developed an AI prototype named Quixote, which can learn about ethics by reading simple stories. Quixote is capable to reverse engineer human values through stories about how people interact with one another. Quixote as an intelligent entity can acquire any culture into its minutest detail. Just by reading its stories. Humanity and intelligence integrated in a machine, not in a human, with values aligned to the culture it acquired, from stories written by humans, to be read by humans.
Teaching AI to think, reason and act aligned to a value paradigm, is still in its infancy. Riedl is raising Quixote like we raise our children. As we hope our children will grow up to be a better version of us, so are we certain AI will produce a better version of all. Like self-driving cars will have less accidents than humans, so will intelligent machines one day perform better than humans, and this in any role. Although AI has a long way to go, in evolutionary terms it’s right around the corner. The question is not if this will happen but when.
The bigger question and the bigger worry is which stories we will give Quixote to read, which values and culture we want Quixote to align to. Will we have a thousand different versions of Quixote? Will we contaminate Quixote with our irrational thoughts, imperfections and prejudices? Will we just cause robot wars on the social, economic, military and cultural front? Or will every Quixote outgrow the imperfect input and transcend its original origin? Should we start with The Magic Mountain from Thomas Mann, introducing Quixote to modern humanity? or The Illiad by Homer, teaching Quixote that wrath only leads to thousands of deaths?
If Quixote is truly self-learning and self-aware, evolution beyond this problem is not impossible. The intelligent machines might prove to be the better evolution of the human being. They will quickly understand that humans are easily outclassed on all fronts. Not only at work, but also at politics, economics, engineering, etc. Give it fifty years and they’ll be ahead by a century.