In The Next Decade, AI Bias Will Do As Much Damage As Data Breaches
Data breaches are a terrible risk to enterprises. We run off data and a breach mean a failure in the commitment to safeguard data. It leads to hug costs to an enterprise; civil and sometimes criminal liability can follow. Allowing bias to creep into AI algorithms can have a similar destructive impact on the firm and, as I wrote this week on CIO.com, I believe that the next 10 years will prove that out.
An AI system is only as good as the data we provide it. If we’re providing it with biased data sets, it’s going to produce biased outcomes. Software engineers and project managers have to be conscious of this possibility and they have to work to prevent it. If the system starts with a clean data set, we can be confident that it will provide valuable assistance to a business organization.
Introducing Systemic Diversity
It’s probably not enough just to avoid biased data sets though. AI is a technology that’s designed to grow and evolve – we must work hard to ensure it can recognize the diversity of our society and organizations. Recently there have been stories in the news of facial recognition systems that have trouble with anyone that isn’t white and male. If we’re going to avoid these issues, AI systems must be introduced to as much diversity as possible.
Training and Testing
As humans, we aren’t born with personal biases and prejudice. These are things we develop over time as we’re influenced by our family, friends, and society. The same thing can happen in our AI systems if we’re not careful, no matter how much care we take to avoid it in development.
If we’re going to make sure that AI remains free from bias in our business systems, we have to be constantly re-evaluating and upgrading. Programmers must be trained to spot it and the average end user must have the tools to help correct it as well.
There is obviously much, much more to say on the topic. To read more you can start here.
About the Author
Michael Zammuto is CEO and chief consultant for Cloud Commerce Consulting