In this episode, Michael and Christian pull back the curtain on the hard truths of applying AI in the industrial domain. It’s not just about flashy demos — it’s about the messy, practical, and often painful hurdles that trip up even the most ambitious initiatives.
We explore five key pain points:
- IP protection — How do you safeguard proprietary algorithms, data and models when they’re feeding into shared platforms or cloud services?
- Hallucinations — Why does AI still confidently make stuff up, and what risk does that pose when you’re controlling physical assets instead of just chatting?
- Reasoning – how far did we come with AI that is not just memorizing, but actually reasoning?
- Costs — From infrastructure to data preparation to ongoing model governance, the price tag for “just throwing in AI” is far higher than many expect.
- The need for machine-readable data & knowledge graphs — AI only works when your assets, systems and processes speak a common language. We dig into why building that foundation (a ‘graph’ of knowledge) is arguably more important than the algorithm itself.
Together, Michael and Christian unpack stories from the field, call out common anti-patterns, and suggest what must go right before the next generation of industrial AI can deliver real value. Whether you’re an executive, a system engineer, or just curious how AI actually scales in manufacturing, this episode gives you a lens on the un-glamourized part of the journey.
Tune in and ask yourself: is your data ready to speak graph? Are your IP walls strong enough? And do you really know what your AI is doing when it “thinks”?
Podcast: Play in new window | Download (44.3MB)
0 Comments