brave souls rewarded today, page-55

It's comments like this that make me continuously question whether people understand the fundamentals of what a machine-learning algorithm is.

For the final time, I'm not an analyst. I have no interest in day trading either. There are already large volumes of of information that will given you significantly more information on the stock prediction in the near-term, than I can give.

What I'm offering to explain is the challenges that a data annotation company faces over the next 5-10 years.

Do you understand the differences between supervised learning & unsupervised learning? If you did, I highly doubt you'd be concerned about a dominance of unsupervised learning. It's also impossible to train an algorithm without data.

Your question about BigData complicates things further. IMHO, BigData is not a requirement for either ML or AI. It's available, but it doesn't mean that the fundamentals of how algorithms are built has changed. Today they are used together but it doesn't mean that the BigData element is a requirement. Having appropriate training datasets is and always will be a requirement for ML and AI.

In your example of customer service, I assume you mean a solution such as a chatbot. Natural Language Programming solutions like this have existed for a very long time. One of the famous examples was MIT's ELIZA in the 60's. In the 70's we had interactive voice response (IVR) menus on telephones. Over the past 10 years, chatbots like Siri/Alexa etc. have become commonplace. All of these solutions require training data.

For a learning algorithm to keep learning, you need data. I'll provide an example below of what an learning algorithm is, so that you can understand the value of human-annotated training data.