Sedna51

Sarah 💫 @harrietvane.bsky.social [ View ] Jul 10, 2024 at 23:48 UTC

I wrote a case study about a machine learning company working on reading X rays/CT scans, and a HUGE problem with this type of process is how hard it is to find training materials for things that aren’t clear cut. We don’t really need AI that can identify obvious negatives and positives.

3 replies 9 reposts 27 likes

Sedna51 @sedna51.bsky.social [ View ] Jul 11, 2024 at 02:17 UTC

yeah, I am very anti-AI in most cases but I think that's one of the few spots where there's some real promise.

1 replies 0 reposts 2 likes

Sedna51 @sedna51.bsky.social [ View ] Jul 11, 2024 at 02:16 UTC

also the higher ups are /really/ concerned about letting third parties have access to our bigger datasets for exactly this reason! we worked to collect this data and told participants it would be used in specific ways, "training a for-profit AI" is not on that list

0 replies 0 reposts 2 likes

Sedna51 @sedna51.bsky.social [ View ] Jul 11, 2024 at 02:12 UTC

some of the biostatisticians at work are chewing over the idea of incorporating AI/machine learning for some of our bigger datasets, but right now it's on the level of quality control stuff. /maybe/ a little further data exploration, but then that would get passed back to an actual human to check

2 replies 0 reposts 4 likes

Sedna51 @sedna51.bsky.social [ View ] Jul 10, 2024 at 22:00 UTC

0 replies 4 reposts 15 likes