AI and Robotics: From Groceries to Babysitting—What’s Next?

2 Jan 2025 · 0 min

Watch on YouTube

0:17

TJ Walker asks whether household robots could safely babysit or hold a baby. The guest explains that robots train in simulators where unsafe behaviors like dropping a knife on a foot are caught and prevented before they ever happen in the real world.

So what all can it do in the house? I mean, I don't have an infant in the house anymore. But is this going to be the sort of thing in the next five years people would trust it to hold the baby or is there still a chance it, you know, crushes something, crushes groceries, it might not matter that much, but a human being again.
So these are trained in a simulator. We can see everything it's doing inside the computer either from the first person's perspective or outside. Outside we could watch it. If it does anything wrong like dropping a knife on its foot, then we obviously don't train that logic. We prevent it from dropping a knife, its foot or whatnot.
So if it does.
Anything wrong or bad, we just make sure that logic doesn't get put on the actual system. So it's very safe because it's contained in a simulator until you publish the or push the AI model to the real world system.