AI has been around for a long time as a few folks have noted. I used AI in the 80s when all it could do was quickly traverse simple Tree Logic (hierarchical model like if - then - else). What has changed are the models they are using now able to use (ordered lists, decision matrices, correlation matrices, flows, and neural nets - closer to how our brains work). That in combination with Machine Learning has changed the game and the complexity of problems the machines can handle.
Machine Learning comes in 2 flavors: Supervised and Unsupervised.
With Supervised - think of training a child to learn something new, you need to properly classify the problem and train the algorithms to "learn" the right way to interpret the data to make a prediction about something. Each time the machine comes up with results that are "outside the norm", we have to retrain the algorithm to bring it under control.
With Unsupervised - This is a bit like discovery, where we have no idea what the results look like and the machine will produce unexpected results...but will learn from its mistakes and over time improve its ability to produce reliable results on its own. This is the new frontier and this is where most of the disruption will occur with jobs displacement.
Any job that is repetitive in nature can be more easily and reliably performed by a machine. Now that we have the ability to train machines with other machines (where they learn from each other), their rate of learning will improve geometrically. So, we already know about the driverless cars and autonomous machines (physical and virtual - lots of little intelligent bots running around the web and larger ones making complex decisions about our healthcare, public safety, and behavioral analysis). But, you need to step back and look at the operations in large companies to visualize the wave that's about to hit. Big companies are designed for scale and they are built on the industrial model of apportioned work performed consistently and reliably all orchestrated using (typically) centralized strategy and decision making. There are 2 threats to consider: First, the obvious replacement of workers at most levels by machines (combo of virtual and physical). Second, is the loss of centralized control since strategy will be more dynamic and decisions will be increasingly made closer to where work is being performed so you wont' need as many executives running companies (hmm).
But...there is a silver lining. As work shifts away from the large (scale) organizations, it will shift back to smaller (more entrepreneurial) endeavors as the availability of AI, Machine Learning (ML), and Big Data become more easily accessible (democratized). Large companies will continue to make commodities, but these will be high volume at low cost. Individuals and smaller companies will begin creating new things that meet our (uniquely) individual needs. The technology is evolving on both ends of the spectrum, so there will be gaps until we reach an equilibrium. That is where we'll likely see pain and discomfort for many. This is where the government may need to step in and balance the social needs with the capital needs. Think about the voting decisions we make today and the ramifications they will have in the next 10-20 years.