Unlocking AI: How Simulators Enhance Autonomous Driving

So the only way you can really get that down is if you take, for example, a model or whatnot and put it inside of a. Inside of a simulator and change the lighting on it, basically in all conditions, from moonlighting, starlight, you know, flash lights of flashing on it or whatnot. So once the AI learns those, those pretty much all conditions are possible scenarios, then it's going to know what the model is and what it's looking at a lot better, including the physics behind it or whatnot. So it's not completely there yet. I know for a fact Tesla's autonomous driving is not completely there. There will always be outliers that will trick or confuse autonomous driving.

Snowy conditions. Dark, rainy. I was in a Waymo autonomous car just a few weeks ago in Phoenix and in Scottsdale, Arizona, and it did work extremely well. You know, things are dry, a lot of sunshine, things are flat. My perception of robots is the hardest part is still the hand movement, the delicacy, the balance. It does appear to be getting better.