When it comes to AI in industry, it’s all about trust
Everyone knows what it’s like when Siri or Alexa gets a voice command wrong. You ask for “Radio Ga Ga” from Queen and end up getting “The Queen” from Lady Gaga. It’s worth a laugh, but nothing tragic.
However, there are situations where AI has to get it right – every time. Suppose AI takes care of end-of-line component testing for an electric drive, and you get the message that the drive is faultless. Great, because that means the time associated with manual testing can be spared. But what if the AI was wrong? Now suppose that drive is built into an e-bus and it fails. That puts people in unnecessary danger – and that’s no laughing matter. In an industrial setting, AI applications must be 100% trustworthy and leave absolutely no room for doubt.
So what will it take to have complete trust in AI-supported solutions in industry? Two aspects are especially important:
1. Data quality
The first is data quality. Logical, really, as it takes dependable, trustworthy data to build dependable, trustworthy AI solutions. But in regard to data, it does require some skill to separate the wheat from the chaff. Siemens can help. Our engineers know industry processes and machines like the back of their hand; they have domain expertise and can access data from a control system or a quality management system; and they know how data from different sources can be integrated and linked. They also have vast experience with data labelling, which is about teaching AI to distinguish between types of data – much like teaching AI in consumer applications to tell the difference between a cat and a dog. Altogether, this expertise has given Siemens a decisive role in creating AI solutions for industry. Now is the time to start collecting data. Even if you don’t use it today, you certainly will at some point. Therefore, get our people involved early on!
2. Active management
Even more important is to actively manage and operate AI solutions in industry – because nothing lasts forever. Recognizing correlations and detecting anomalies with AI in the proof-of-concept (PoC) phase is one thing. But bringing a PoC to life, implementing it into industrial IT and OT, and keeping it operational is even more challenging. A reason is that manufacturing environments are constantly being adjusted – products change, suppliers are switched, parts wear out, sensors get clogged, and operators modify settings. You have to be aware of changes like these – and then take necessary steps to ensure AI solutions continue to operate as they should. We at Siemens can help in the framework of our Managed Services. You win through long-term benefits from your AI solution – especially when it’s constantly monitored and operated by us!
AI solutions for industry require trustworthy data and quality care. Only then can they really add value. As a pioneer in industrial AI, we at Siemens leave the hit-or-miss mentality to the others.
Now time to get back to my playlist. Let’s see what Spotify suggests – sometimes I have the feeling it knows what I like better than I do …