For seamless interaction, a system needs to have precise knowledge about the activity and the habits of a users so it can support them in an adequate manner. The system needs to have knowledge about objects as well as situational observations and historical data. This input is not something that is done by a keyboard, but is done through the use of sensors. Different sensors observe different kinds of information, even to the extend that exceed human capabilities. The problem however is not the registering of information, but the processing of this combined data streams, creating input of a rich context. As humans we are very selective to all ‘incoming ‘input’, we could absolutely not function if we would not prioritize and thus filter this input. We can only come to a decision when we filter the information. For a computer today this is still very hard to do. Especially when it comes to making a judgement based on different environmental information as it often includes a lot of ‘noise’
For a computer picking up a word as a command is relatively easy to do. But when this word is embedded everyday noisy environment is gets a lot trickier.
The logic behind these systems are becoming more and more complex. The immediate problem is not the registering and transfer of all sorts of input to networked computing devices, the difficulty lies in the connection of gathered logic and the filtering and interpreting of information and noise to finally create a judgement that needs to be done before executing a task independently and properly. Take a look for example at the effort that is being put into the autonomously operating cars in de Darpa Challenge. The amount of sensors that are connected to the operative unit of the vehicle is quite large. And this is ‘just’ for driving the car independently from human. By the way, most of the cars that participated in the contest never crossed the finish line.
Another field that is quite interested in automated processing and interpreting a numerous feeds of visual information is Las Vegas. With it’s millions of visitor and the daily amount of cheaters, the casinos like to have autonomous cheating detectors. Again, the problem with this lies not in the incoming data streams in their multi million dollar security rooms, but in the processing of this information into judgment. The power of human eyeballing is still far greater than its autonomous sensory counterpart. The difference in judgment between a man scanning for fruit machine fiche’s and a man looking for his family is still far too subtle for a automated devices to detect.
