There are a few different reasons why people are talking about the “tipping point” of machine learning and how this type of technology will apply to future applications.
First, as the global community has had time to learn about machine learning, we see that current technology in neural networking and various types of algorithm development support the key idea that machines can be self-taught or “learn” over time.
Free Download: Machine Learning and Why It Matters
A few years ago, by contrast, machine learning was still brand new. Many people did not know what it meant, and applications were scarce.
Now, with more developed processes for building machine learning programs, aggregating the data for training and testing, and importantly, applying conceptual ideas to enterprise, there's much more of a sea change toward machine learning and related artificial intelligence use cases than there was a few years ago.
In a way, the move toward machine learning and artificial intelligence is similar to the dramatic change that business went through in the past two decades, as cloud computing modernized data storage and the use of the Internet.
Phenomena and trends like cloud computing and the Internet of Things have revolutionized networking in many profound ways. Machine learning and artificial intelligence are now doing very profound things with our concept of how technology applies to a knowledge base. Experts might point to other related specific drivers for ML advances: for example, better strategies for allowing ML programs to converge, better setups for pipelines to deliver ML functionality for specific use cases, and a larger field of talent being created by schools and tech training institutions.
On another related point, another driver of this tipping point with machine learning is hardware evolution. Since machine learning technologies tend to need certain kinds of processing power, the most modern chips and microprocessors have been made to accommodate that. Quantum computing has also arisen as a solution to big data problems around some types of machine learning endeavors.