I am not an artificial intelligence (AI) (NLP, Machine Learning, Deep Learning) expert, but I am a Cognitive Psychologist with a keen interest involved in developing a deeper knowledge and skillset in these areas. As a researcher with knowledge of human factors (human abilities and limitations related to human sensory systems), I asked the question, “Do artificial intelligence strategies, approaches, algorithms, and applications, fully and appropriately utilize human factors in their development and modelling of human-like tasks?” It is my opinion that at least the classic and cited examples of supervised learning does not.
This point of view is derived from many textbooks, online courses and tutorials, and training I have had and/or experienced in/on supervised learning. I will use examples on supervised learning from the May 22, 2019 Forbes Insights article, titled “What Machine Learning Needs To Learn Next” (Forbes, 2019). This is a great article which covers some of the ways both supervised and unsupervised learning is used. The author (N.D.) purports that “most machine learning uses a technique called supervised learning.” An example of supervised learning is described where a cat is being distinguished from thousands of pictures of cats and pictures that are not cats. Supervised learning algorithms were trained using thousands of pictures that were found and labeled for this purpose.
I think that is similar to the logic expressed by Bernard Marr in a June 3, 2019 Forbes article, titled “5 Amazing Examples Of Natural Language Processing (NLP) In Practice (Forbes, 2019). When discussing this form of AI, Bernard Marr states that “every day, humans exchange countless words with other humans to get all kinds of things accomplished. But communication is much more than words—there’s context, body language, intonation, and more that helps us understand the intent of the words when we communicate with each other.” Natural language processing is used in a wide range of applications, but realizes the importance of human factors to accomplish its tasks.
Now back to the supervised learning examples. For the distinguishing of a cat from thousands of pictures (cat and non-cat), how efficient and accurate would the assessment be, if it this case, human-like factors such as anatomy, head size, head shape, body size, body shape, spatial location and position, context, gestures, actions, etc. were combined with contrast, color, number of pixels, and other normally used output characteristics in the algorithms? Would you then find distinguishing a close relative like a tiger, or dog difficult and require more training and test sets?
References:
Marr, B. (June 3, 2019). 5 amazing examples of natural language processing (NLP) in practice. Retrieved from https://www.forbes.com/sites/bernardmarr/2019/06/03/5-amazing-examples-of-natural-language-processing-nlp-in-practice/?ss=ai-big-data#d0cb5561b305
Forbes (N.D.)(2019). What Machine Learning Needs To Learn Next. Retrieved from https://www.forbes.com/sites/insights-intelai/2019/05/22/what-machine-learning-needs-to-learn-next/#b6f9a56a8899