AI. Machine learning. Outside of tech, these things instill anxiety, because people don’t understand their implications. But those in the industry can sit at the other extreme, attempting to infuse machine learning into everything they do. Neither approach is good.
Mark McQuade, Practice Manager, Data Science & Engineering at Rackspace Technology, takes a more measured approach. As someone who works on “all things data — all things machine learning,” his thinking is distinctly more scientific in nature.
To win over anyone wary of machine learning, McQuade suggests pointing toward the countless practical and positive ways it currently impacts businesses and individuals alike, including prediction, classification and recommendations. “Machine learning does so much today that people don’t realize they’re constantly getting benefits and value from it,” he said. For example, the more human angles of machine learning escape people who are unaware of sentiment analysis, which ascertains emotional states from live call center transcripts, bulk online reviews and apps that are geared toward helping people with disabilities. Through machine learning, all of these things occur at scale to a point that would be impossible for humans to match — even if backed by powerful traditional computer programming and logic.
Our latest episode of the Cloud Talk podcast explores the many sides of machine learning.
In just 30 minutes, McQuade and Rackspace CTO Jeff DeVerter explore:
- How AI and machine learning are related
- The basics of machine learning and how it can bring value to an organization
- Why ‘AI for good’ principles should drive your adoption of machine learning
- The complexity of AI ethics and machine learning guiding principles — and why they must be considered on a case-by-case basis
- Why only 60% of machine learning models see the light of day — and whether MLOps can fix that
McQuade and DeVerter also address one other big question: Should we use machine learning at all? “Many people just want to be involved in that world and get machine learning into their organizations,” said McQuade. “But they must ask why, because not every problem is built for machine learning. You must judge whether what you’re trying to accomplish is complex enough to warrant machine learning in the first place. If you can achieve something using computer programming and logic, there’s no need to make it overly complex with machine learning, which could hamper your ROI.
With certain tasks, though, machine learning is the best choice. “You can automate processes to make life simpler for your teams, better your organization and focus staff on what adds value,” said McQuade. “This goes beyond use-cases like manual repetition: For example, object detection via basic computer programming and logic is hard. So whether you should use machine learning isn’t solely down to something being repeatable. It’s about how hard something is if you don’t use machine learning compared to if you do. This isn’t only about simplifying processes — without machine learning, some things you want to achieve just won’t be feasible.”