It's fun to see everybody succumbing to scaling laws. We will only have FSD when we can run realtime inference with a huge model. No amount of creativity seems to overcome the fact that bigger is better. So, let's wait Moore's law do its job.
That's fun. We should join their AI team.
I’m not sure what architecture they use, but they do indeed already have a pre trained “auto-labeler” that their annotators use. My understanding is that due to hallucinations from the model and the risks involved with driving, they still need to be vetted manually before being added to the dataset.