I’m actually glad to see this article since I’ve been writing about this topic for nearly a decade and it seems the design industry is stuck on interaction design. It is an interesting article though I think it shows the difference between researching a topic and actually executing. I would argue that design thinking is even more apropos for designing with AI since I’ve been using it for more than a decade to do just that because it is a good methodology for exploration which is what’s needed for a stochastic design outcome. Method such as speculative design, design fiction, Wizard of Oz, are all great for designing with AI and even more necessary because of the uncertain outcome. I also think there’s a huge important phase missing here called UX Research that is integral if not more important than design when it comes to enabling AI-powered human-centered experiences because that’s where your blending of data science with design research can influence the model outcome more so than any design.

And you don’t need less prototyping you need more and earlier; precisely because you have no idea how people will react to AI outcomes or how those outcomes will effect the experience. Co-design and co-participatory design workshops help you gain more clarity and certainty.
And really, in my experience, the biggest difference in designing for AI isn’t it’s probability (good simulations can take care of that and provide more certainty) it’s the ability of an experience to evolve. Which is why I speak in terms of the human- machine relationship because when designing AI-empowered experiences you’re designing something that will change overtime and this is the second paradigm shift. The first is designing for a stochastic or dynamic outcome. Prototyping is a method that can help you deal with both challenges. Why would you ever not do that?!

But I do agree that understating capabilities and limitations of models (we can say AI but most likely you’re talking about machine learning models working together to achieve a prediction - AI doesn’t happen until you add IoT and contextual ambient sensor which allow machines to make real-time decision making ) is a great start for designers in this space. It will help you to grasps the cognitive paradigm shifts needed to successfully transition to AI design. I share those in my talk Design in an Automated Future.
And while I know it’s all the rage for folks to spend months reading about something that has been in existence 30 years and instantly declare it irrelevant let’s not throw out perfectly good design methodologies just because a technology has gained new capabilities. Human-centered Design is it a temporary framework or a quickie online method it’s a profession and it’s just as sound as science is for quantum computing it can handle AI.