GPUs a ‘Game Changer’ for Nuking Noise in Nuclear Imaging
blogs.nvidia.com
blogs.nvidia.com
Deep networks can learn to hallucinate rather well. I'm sure that they can fill in what body structures are supposed to look like (they've seen them many times) but not sure if they would help in diagnosis -- or if they would add plausible-looking details that would breed complacency.
IMO this kind of astroturfing shouldn't be allowed without some kind of disclaimer even if the content is pretty interesting. There should at least be something in his bio about it.
But it is advertising and it is not good for communities at all.
It's ok on HN to submit your own stuff as long as it's mixed with a variety of other material that you ran across and personally found interesting.
Using HN only to submit own material is less cool, but not something we necessarily ban users for. The community tends to frown on and flag such submissions though.
bcaulfield is obviously a legit HN user. The proportion is probably too heavily on the Nvidia side, but that doesn't mean he should be hounded off HN either.
Hopefully the probabilities of error are carefully quantified.
Edit: speeding up the MC estimates of absorbed dose, etc, (i.e. speeding the treatment planning process) seems like a much more clear improvment.