From what I have seen, the biggest shift is not perfect generalization yet, but better perception and learning from real-world data. The demos look more capable, but reliability over long periods is still the hard part. It feels like progress is real, just slower than the funding hype suggests.
From what I have seen, the biggest shift is not perfect generalization yet, but better perception and learning from real-world data. The demos look more capable, but reliability over long periods is still the hard part. It feels like progress is real, just slower than the funding hype suggests.