Rain, neural networks and the promise of AI

During a recent speaking engagement, I highlighted the pitfalls of viewing AI as a “catch all” —a one-size-fits-all solution to every problem.

While AI holds transformative potential across numerous sectors, it doesn’t mean we should shoehorn it into every application.

To illustrate my point, I shared a personal anecdote. On my drive to the venue, it started to rain. As the rain obscured my vision, I found myself having to manually activate the wipers. Why? Because the car’s AI-driven “deep rain” system—which supposedly uses a neural network to analyze camera feeds around the vehicle—failed to function properly. Ironically, most times, I end up manually operating the wipers due to this very issue.

This feature, which has been in development for nearly four years (and probably more), consistently underperforms. A quick google search shows that traditional rain sensors, like the ones in Volvo models, cost around 50 Euro (on consumer market).

It makes one wonder: Was the “Deep Rain” approach truly a cost-effective or practical choice? I know what i think!

Think about this example when you evaluate your next use case. 🙂