Anyone who's ever attending an industry talk knows the Q&A sessions generally are pretty benign. But once in awhile it can take an interesting turn.
That happened a couple months back when I attended the International Institute of Synthetic Rubber Producers Annual General Meeting in Seattle. Microsoft's Egbert Schroer gave an interesting talk on IT Developments & Mobility, touching on such topics as open manufacturing platforms and digitization across the value chain.
The Q&A session was conducted using Slido, where attendees submit questions via their phones or tablets—either anonymously or with their name attached—in hopes of increasing event interaction.
But one attendee stepped forward to a microphone. She declared she was "old school," and she could ask her question quicker live than by typing it into her phone. Her job involved being a troubleshooter, and she found that once young engineers see data pulled into a report, they take it as gospel. "And then you have to spend your time to say just because it's digitally put in a graph doesn't mean it's real," she said. "So my question is whether Microsoft is teaching them to think and say, 'You know guys, this is out of normal ranges. If it's doing this, that's not normal.' "
Schroer welcomed the question, saying he was old school as well. But with all the benefits of artificial intelligence, there still needs to be a way to differentiate good data from bad. Microsoft has had conversations with customers about the topic, and the answer almost seems a bit counterintuitive.
"AI can only be as good as the human is," Schroer said. "So you need to train the system and how to train the system depends on—what a surprise—the experts. So the experts need to tell the system what is real and so on."
Another potential help could be a course in common sense for the young engineers the questioner said couldn't think for themselves.