Intelligence in a Box is Wishful Thinking
The idea that software can be intelligent is widely endorsed by scientists and laypeople alike.The clearest manifestations of this belief were the so-called expert systems which during the eighties experienced a global boom.The goal was to extract and formalize “intelligence”from experts so that it could be stored in a computer and made generally available, such that anyone could tap into this precious knowledge.The concept gained new momentum with the advent of the Internet:After all, the amount of potential knowledge at our disposal there is mind-boggling. But the failure of expert systems in practice, corroborated by theoretical investigations, has demonstrated that these ideas are based on an inappropriate understanding of intelligence, knowledge, and expertise. What is stored in a computer and the Internet is data, not knowledge:The data has to be transformed into knowledge. Knowledge is what guides our actions; it always depends on a specific context which is defined by the current goals and the current real-world situation of the agent. It cannot be stored in a box stripped of context.
Rather than trying to store intelligence in a computer, we should aim at understanding the interactive contexts in which individuals make decisions and how the intelligence of agents can be enhanced. It is the hybrid human/computer system that bears the greatest potential.