Machine Learning has three main features.
For some, recommendation engines one websites are valuable tools, perhaps suggesting good movies similar to ones you've watched already, based on preferences
For some, clustering of homes on real estate sites are good tools. What is my home value based on similar homes within the region.
And for others, classifying images is a good tool, perhaps recognizing your friends in a picture on social media, without having to enter the tags in manually.
Recommendation, Clustering and Classifying. Three solid pieces of machine learning or types. And a variety of algorithms to get to that goal.
There's also Neural Networks, that are useful for artificial intelligence, they have weights that get toggled true or false, depending on incoming variables, with multiple layers, which produce an end result weighted probability.
Some common themes are lots of data, to train the model, to learn from known data. some data needs to be labeled ahead of time, which is "Supervised Learning". The opposite is having the model learn on its own, with the luxury of "tagged" data ahead of time, called "Unsupervised Learning".
There're also models that learn from other models, to reinforce learned behavior, the event happened, this was the result, let me track that, for future reference.
Some Neural Nets are very large, with hundreds of layers deep, to get fairly precise results. They do require more compute power and memory and take time to process.
There are a slew of languages and tools to use when working with machine learning and artificial intelligence, both on-premise, virtual machines and in the Cloud.
The fact we have bigger data sets, better compute power, more ram and machines chips that can crunch more data faster. In the past, typically large institutions had access to these types of machines, so Universities or large computer organizations were the only place to work on this technology. Now you can program this on a laptop in you living room if you desire. So things have tricked down into the hands of the many. Thus, faster progress, cleaner solutions, better results.
The results have gotten better, in image recognition, speech recognition, translation tools from language to language, some in real time and many more. At this point machine learning is a useful tool to assist humans in everyday activities. It has not gotten to the point where AI is replacing everyday jobs, in most cases. But the tides will shift at some point, where people are competing with smart machines for everyday jobs. And when robots are tasked with specific tasks, that too will automate some of the workforce. And when smart machines can perform complex tasks in real time with minimal error, then we will take notice as computers do not necessarily require vacation time, or health insurance, or 401k matching contributions, and they work 24/7 without complaining or forming Unions.
So, we have come a long way, computers can crunch data and translate into meaningful information for consumption, to lower costs and automate menial tasks. Where do we go from here? More automation, easier to use tools, better integration across domains, access in everyday tasks, embedded IoT devices that assist in real time. Artificial Intelligence should not be interpreted strictly as computers rising up, taking over man kind. Surely, anything is possible, at this point in time, we are still telling the machines what to do. And they do their specific tasks, in a sort of black box, and we do not necessarily know exactly how the results were derived.
With Moore's Law, things will only get faster, cheaper, easier to use and proliferate through everyday society. That should keep us busy for a while, for sure.