In the second episode of our series on AI we will look at Machine Learning which has been very hyped recently, and is often confused with AI. Remember from our first episode that AI is an algorithm that performs a cognitive task. Machine Learning is a class of algorithms that perform a cognitive task (classifying items into groups) with the classification rules being created by examining past data rather than by a programmer. Because Machine Learning algorithms are conveniently available, many problems that you might not think of as classification problems, are defined as classification problems to use these algorithms. For example, food in a picture could be defined as Hot Dog or Not Hot Dog, facial recognition could be classified as Paul or not Paul, and if not Paul then John or Not John and so on.
There are many types of machine learning algorithms and whenever a new one emerges, that advances a particular problem, it gets hyped. Recently, a class of machine learning algorithms known as "Deep Learning" have advanced our ability to recognize elements in images (which helps solve many problems such as facial recognition and obstacles in the way of self driving cars) and recognizing words in spoken audio. In medicine, machine learning is trickier.
A major challenge for using machine learning in medicine is reproducibility. Algorithms that constantly change their behaviour need to be constantly monitored to ensure they don't get worse over time. That means turning every use into part of a trial and most hospital systems don't have the capacity to do that yet.
Machine Learning Takes Heat for Science’s Reproducibility Crisis
Leave a Reply
Comments, questions or queries? Let us know!"