Last year, my predictions for artificial intelligence (AI) had an understandable technology bias—after all, AI is all about the tech. But this year, my predictions focus not on the what, why and ‘can we’ of AI, but on when and how it should be used. These will provide the two overwhelming AI trends for 2020.
Debating the ethics of AI
There is already a lot of activity and debate around the topic of ethical AI. This will grow in 2020. The ethical challenge organizations face is one of both regulatory compliance and customer trust. The compliance angle is easy to get to grips with—stringent regulations, most notably the EU’s GDPR, restrict the use of certain sensitive data, such as customer data. Those organizations using actual customer data risk heavy financial penalties for non-compliance. Thus, there is a strategic imperative to use and store customer data correctly and ethically.
At a trust level, we all know that organizations capture data from their customers in order to better respond to and serve those very same customers. But just how much data do they need, or should they actually use? Can consumers trust these organizations to use it ethically? As AI makes it possible to go to extreme lengths of personalization, this debate takes us beyond what I described last year as ‘explainable AI’ and into the sphere of personal privacy—a basic human right.
How far do new and emerging AI use cases take away that right? For example, could AI be used by insurers to predict life outcomes? What happens to imagery of children and parents captured by AI-enabled entry systems/doorbells at school gates? These are among the ethical questions all organizations pushing ahead with their adoption of AI will be considering in 2020 and beyond.
The rise of synthetic data
If regulation or ethics prevent you from using real customer data, how do you create data sets on which to build a differentiating customer experience? This question will drive another big trend in 2020—that of synthetic data. This is data that’s synthetically generated out of your original data set using deep learning methods. This synthetic data set closely matches the original data in terms of statistical similarity and distribution and can thus be used in place of the actual data. The data set can then be used to help build a better AI model with no compliance risks. This is a topic that Sogeti has explored in depth, enabling us to develop a solution that is already helping several clients to create synthetic data for all manner of applications.
Get in touch to find out more about our solutions incorporating AI and synthetic data.