|What's wrong with Machine Learning?|
What’re the issues with Machine Learning?
There are 3 major issues and their solutions are not in sight. However, these issues have not stopped Machine Learning to reach an inflection position at which commercial organizations are subscribing huge resources to develop Machine Learning in parallel with academic researches. The same info here was publicly discussed in a Parallel Computing Meet Up gathering dated 2019-04-03 in Compucon House.
(a) Machine Learning is capable of achieving a higher level of accuracy than traditional computing algorithms (such as linear regression, support vector machine, and random forest). Why is it possible? Why can’t the traditional computing algorithms be improved? Some Machine Learning specialists claimed that machine learning mimics how the human brain learns. Is it true? Do we really understand how the brain works? Not really at the moment. If we do not know how Machine Learning achieves the accuracy being demonstrated, can we trust Machine Learning to do jobs that may affect health, safety, security, and so on?
(b) Machine Learning is very specific to the data from which it learns. This is totally understandable as young children learn from their parents, school teachers, and fellow pupils. Well yes it is understandable but it can be seen that Machine Learning is not as well exposed as children in the learning phase. A consequence is that it is not easy to adapt a trained Machine model to a new set of data. In contrast, children are quick to take on new information and knowledge.
(c) Machine Learning needs a huge amount of data for learning and a huge amount of computing time and resources for training. This implies that organizations with access to a huge amount of data and computing resources are in a better position to do Machine Learning. Creativity alone is not sufficient. Scale is needed.