three distinct questions to ask that would distinguish this:
Professor Jaakkola went on to say that even if you can formulate a machine learning problem, and have tools available to assist you, some problems are inherently impossible to solve, especially if you are trying to predict a chaotic system. One such example that he provided was planetary motion.
Using the three questions outlined above as a guide, think of some example problems in business or elsewhere. Which of these problems are solvable with machine learning? Which problems would be impossible to solve?
There a couple of " problems" which generally are impossible to solve using Machine learning, E.g. can machines solve world wide poverty ? This problem can never be solved by machines as this can be termed a "chaotic systems" as there are multiple diverse input parameters which could consist of economics , cultures and environments etc etc. In the same context, machines cannot resolve the "Global warming" or reduce global green house gas conditions of the world. Hypothetical, this would be possible if we have terraforming at a planetary engineering process level. currently we just don,t have the tools for such a concept. Where we can use machines to objectively alter the environment at a a global scale.
On the other hand, we have break through's with Machine's learning to analysis mammograms images and identify if the patient has breast cancer. The Machine learning design was using Convoluted Neural Network (CNN). The results from the machine are much more accurate then a human diagnoses. Also the output/ results are much faster than humans beings. So the machine is actually saving lives !! Another concept is heart attacks predication