“The horse is here to stay but the automobile is only a novelty---a fad.”¹
There is an inherent risk when making a prediction of the future that you will inevitably look very silly in a few years. But it’s also a fun end-of-year tradition to think about what the coming year holds. Without further ado, here are three predictions for where artificial intelligence (AI) will go in 2019.
Is It Still All about the Data?
As the saying goes: garbage in, garbage out—and the same holds true with artificial intelligence. In fact, AI without good, trustworthy, and governed data to build models with is exactly that: artificial.
In 2019, we’ll continue to see an evolution of how data is stored, managed, and governed. In particular, there is a need for a content management system for your data. For example, if you use iCloud or Google Photos, you can have pictures from anywhere in the world, stored on multiple devices, but easily searchable and consumable.
The same will become applicable for data: be it structured, unstructured, audio, video, image, or text. And AI teams will need to manipulate all this data so that it can be used to build AI models. However, all that hard work should be preserved so that the next AI team can find and reuse these datasets with ease.
Scale—Not Just for Dirty Faucets Anymore
AI systems can help fix things before they break, they can help you understand customers better than ever before, they can even talk to your customers on your behalf. But when the cost to develop, deploy, and maintain these systems is too significant, their prevalence in the real world becomes limited.
If it costs you $5 million to build and create an effective AI model, you will only solve problems that cost you more than $5 million,. There are a limited number of $5M+ problems in the world, and many smaller problems that can benefit from AI techniques.
The way to address this is though automation—scale requires that you find efficiencies and remove low value repetitive tasks from the equation. Intelligent Robotic Process Automation lets you do this for your knowledge workers in key applications,and intelligent interfaces and conversation AI allows you to do this for man-to-machine interaction. But we will see an increase in solutions that allow you to do this for the underlying process of designing, deploying, and maintaining AI models.
The AI Assembly Line
This is nothing new—the first and second industrial revolutions were driven by the assembly line. By breaking down big jobs into smaller tasks that could bring repeatability, specialization, and automation, we ensured that cars became affordable, that goods and services could be within the reach of a broader class of society, and that they could be easily transported to anywhere in the world.
The same will happen with AI when you fundamentally need to connect to the data you need, learn from that data what the signal is, scale the application of that across your organization, and ultimately consume that knowledge either through an automated business process, a next generation user interface, embedded in an application, or directly in a data visualization.
What are your thoughts and predictions for how AI will evolve in 2019?
Learn More
¹Infamous “worst prediction ever” from a banker to Horace Rackman, one of the early investors in the Henry Ford company.
