Machine Learning systems are all set to surprise us and here is how. The simple spreadsheet of machine learning foibles is a fascinating exploration of how machines think. These machines obviously are not capable of thinking in the real sense.
When it comes to a given set of parameters and the ability to evolve an answer, it becomes but obvious that these robots will come up with various fun ideas. There was a test wherein a robot learned to move a block by simply smacking the table with its arm.
In another example, a cancer detecting system found that pictures of malignant tumors usually contained rulers and hence gave plenty of false positives.
These examples depict the unintended consequences of trusting machines to learn. Machine learning basically means learning that is understandable only by machines.