Artificial Intelligence in 2018
It’s that time of year again! What can we expect from Artificial Intelligence in 2018? Read on for my predictions for the artificial intelligence (AI) trends we’ll see in 2018.
AI Prediction #1: The fog around AI will start to lift
The hype that has surrounded artificial intelligence for so long is being recognized for what it is, allowing us to accept and even embrace certain realities of working with AI-based technologies. In 2018, I believe that organizations will understand that automatic learning IS NOT REAL.
Training a machine learning-based system requires an investment of resources, time and the patience to endure the many reiterations required of any system that has to “learn.”
AI-based automation of customer care and other back office process provides also access to more data, especially unstructured data. With supervised machine learning, which is required in most situations but especially with unstructured data, organizations need very large data sets in order to provide automation for enterprise processes. This data must also be labeled so that the model can “learn” to identify the correct outcome.
For even a simple model, labeling might require 35 seconds per label; if you have 100,000 pieces of data, that amounts to thousands of hours of work. This doesn’t include the reiterations typically required for reaching acceptable outcomes
Once this reality sets in, the radical acceptance model of machine learning as the only technique for applying AI begins to fall apart. Instead, the more realistic (and easier) way to do it is to mix AI with other more traditional techniques based on systems that have a body of knowledge, that “understand” and, with some adjustments, can be adapted to work in complex situations.
AI Prediction #2: AI will not save us from fake news
Although the most prevalent distributors of fake news, the social media platforms, claim to be committed to fixing this issue, the economic and political incentives are too large for that to happen just yet. In addition, new applications of AI, such as with Virtual Reality for example, will only boost the development of more advanced forms of fake news. Addressing the phenomena will require a global effort, but I don’t see this happening in 2018 (nor in the next 5 years). Organizations, individuals and society at large will have to have to find other ways to deal with fake news for the time being.
AI Prediction #3: AI as insight systems
Intelligent apps are increasingly adopted to primarily augment, but also replace, human activity. The automation of customer care and back office processes is making a new class of unstructured data available. The content in communication streams, such as in supporting documentation for customer requests, provides a real time-view of what is happening in an organization’s relationship with customers. Using AI techniques for automating data preparation, insight discovery and insight sharing will provide more support for decision making and research to a broad range of business users, operational workers and citizen data scientists.
AI Prediction #4: Digital Models
A growing trend in 2018 will derive from our access to increasingly sophisticated digital representations of most aspects of our world. AI-based capabilities will help make it easy and relatively inexpensive to perform advanced simulation, operation and analysis. For asset-intensive sectors like oil and gas and heavy manufacturing or for digital marketers or healthcare professionals, it will open up a virtually unlimited amount of possibilities for testing new scenarios. For example, future digital models could provide biometric and medical data, and digital models for entire cities will allow for advanced simulations.
AI Prediction #5: Understand the ROI of AI
Organizations will have to think about how they calculate the return on investment (ROI) for AI. There will be two kinds of measurements. Traditional measurement related to efficiencies achieved through automation (such as for deflection of customer calls and for back office automation) and revenue growth will be augmented by a second kind of measurement. This would include, for example, automated full-time equivalents (AFEs) as automation takes on discrete tasks or business processes and frees up human workers, or less tangible returns, such as increased confidence in an outcome or the additional options an AI solution provides. This second kind of measure is more difficult to calculate, but it is also gets us closer to the real value of AI. Ignoring this (and postponing your AI investment) could directly impact your future ability to compete.