When humans process information, it is understood that not every piece of information is created equal. For example, if in the world of finance a regulator, a company’s management, some reputed analyst and some infrequent blogger comment about something, humans will automatically rank order the various statements and get to the truth much faster. It also makes is easy to process contradictory pieces of information.

Currently AI does not behave like that. For a typical neural network, all this information is equally important. Not only does this make training much more compute intensive, contradictory information goes to dilute the models rather than strengthen it. In other words classical AI models are unnecessarily more expensive, less accurate and harder to program when dealing with real world data.

Out tactical cognitive computing platform uses the concept of Cognitive Calibration to better emulate the human learning process. We speak more about it at the With The Best Conference on September 25th, 2016. Make sure your tech teans attend the session.