What is a generalisable model? A model that can adapt to previously unseen data. How does that translate to self-driving? An AI driver that can adapt to any vehicle, and drive in any city, in any weather. What does that look like in practice? 👇

It ran over the roundabout?

Impressive but needs a little fine tuning. Ran over a roundabout and then drove far too close to the cyclists.

Intriguing but I am skeptical. It seems like this approach would require way too much data to get to 99.9999% reliable plus there so many edge cases. You can probably do 90% of driving this way but any edge case that was not in the training set will trip up your car.

This country needs a full rewrite of their roads. It’s horrible in every possible way

@memdotai mem it

@wayve_ai Saved! Here's the compiled thread: 🪄 AI-generated summary: "A generalisable model is an AI driver that can adapt to any vehicle, and drive in any city, in any weather, and this is demonstrated in practice through the link provided."

@get_this_v

You haven't seen Medan yet. Need to train a specific model for this city. Nothing generalisable about its traffic.