Technology

I totally agree with you, but that problem is way harder to solve than LLMs and music generation.

We got LLMs and music/image generators first because there's so much written information, music recordings, and images to train models on. There's practically zero data about how your body moves when you're washing dishes and what it looks like in the sink, what a dirty plate looks like or a clean one, so we can't just train robots to do it yet except in extremely specified environments with as few variables as possible.

There's a lot of advancement in this area, trying to solve these problems. Robotics is a very exciting field of advancement these days. Automated dish cleaning machines exist that don't use AI, but they're industrial and expensive. If you want something that uses dawn, a sponge, elbow grease, and your particular style and height and form factor of kitchen sink, we need a lot of data about people washing dishes to train models on.

At least, at this point we do. We will probably get to a point where an AI enabled robot can learn new manual tasks, but think about how long it took YOU, a human, to develop and learn motor function and coordination. All that motor telemetry data, input sounds, sights, and tactile sensations are handled by your brain silently, and we take it all for granted. But it's all built off a lifetime of your brain taking input data and processing it so you can run the biomechanical series of squishy meat levers and pulleys to wash the dishes.