Re:“AI promised to revolutionize radiology but so far its fa...
by mark_l_watson on Monday June 07, 2021 @11:23AM
(#27423651)
We used AI to analyze seismic data in the DARPA nuclear test monitoring system in the 1980s. I don’t think that it was considered to have anything but a fully automated system. That said, we had a large budget, and great teams of geophysicists and computer scientists, and 38 data collection stations around the world. In my experience, throwing money and resources at difficult problems usually gets those problems solved.
[ Parent
| on HN ]
Re:“AI promised to revolutionize radiology but so far its fa...
by jofer on Monday June 07, 2021 @11:28AM
(#27423716)
Very different sort of seismic data, FWIW.
You're referring to seismology and deciding whether something is a blast or a standard double-couple earthquake. That's fairly straightfoward, as it's mostly a matter of getting enough data from different angles. Lots of data processing and ambiguity, but in the end, you're inverting for a relatively simple mathematical model (the focal mechanism): https://en.wikipedia.org/wiki/Focal_mechanism
I'm referring to reflection seismic, where you're fundamentally interpreting an image after all of the processing to make the image (i.e. basically making a mathematic lens) has already been done.