So on appoint! Even the god father of AI (Geoffrey Hinton) is wrong about this. Yann LeCun, who is his student way back also isn’t afraid to point that out as well.
Exactly, there are a whole hosts of ground models that are not explicitly expressed in text or images that buried deep inside medial professionals brains. It’s a far stretched extrapolation from some image classification model to it will demolish the radiologist profession in 5 years. Eventually, yea, but we need much better learning techniques, and medical data collection and collaboration (somewhat even harder... due to PPI, regulations, and such)
That’s been said, there are some low hanging fruits that can improve efficiency and lives of the medical professionals in general not just radiologists. If you ever walked into any family doctors office or unfortunately enough to be hospitalized, you’d see doctors always carry a clipboard or sitting in front of computer typing away or they are followed by a flock of residency doctors with clipboards or tablets and they barely look at you, the patients. Because they have to keep records as much as they can for references, for patient records, for evidence purposes should anything happened. The existing LLMs with some enhancements (PSFT+VectorDB+Langchain,etc.) can create an easy to use app to help them record those data(audio, text, image, video) which can be used for further knowledge training, reduce their filling out form time, and protect them from malpractice or unreasonable complaints (which happens a lot in China...).
So on appoint! Even the god father of AI (Geoffrey Hinton) is wrong about this. Yann LeCun, who is his student way back also isn’t afraid to point that out as well.
Exactly, there are a whole hosts of ground models that are not explicitly expressed in text or images that buried deep inside medial professionals brains. It’s a far stretched extrapolation from some image classification model to it will demolish the radiologist profession in 5 years. Eventually, yea, but we need much better learning techniques, and medical data collection and collaboration (somewhat even harder... due to PPI, regulations, and such)
That’s been said, there are some low hanging fruits that can improve efficiency and lives of the medical professionals in general not just radiologists. If you ever walked into any family doctors office or unfortunately enough to be hospitalized, you’d see doctors always carry a clipboard or sitting in front of computer typing away or they are followed by a flock of residency doctors with clipboards or tablets and they barely look at you, the patients. Because they have to keep records as much as they can for references, for patient records, for evidence purposes should anything happened. The existing LLMs with some enhancements (PSFT+VectorDB+Langchain,etc.) can create an easy to use app to help them record those data(audio, text, image, video) which can be used for further knowledge training, reduce their filling out form time, and protect them from malpractice or unreasonable complaints (which happens a lot in China...).
Great comment. Please what's PSFT?