[ad_1]
Artificial intelligence is used with excessive haste to badyze data in some areas of biomedical research, which has led to inaccurate results, warned Friday a prominent US computer scientist and medical statistician.
"I would not trust a lot of the discoveries currently being made using machine-learning techniques applied to large data sets," warned Genevera Allen of Baylor College of Medicine and of Rice University at the annual meeting of the American Association for the Advancement of Science.
Machine learning is a form of AI widely used to find patterns and badociations between scientific and medical data, for example between genes and diseases. In precision medicine, researchers are looking for groups of patients with similar DNA profiles so that treatments can target their particular genetic form of disease.
"Many of these techniques are designed to always make a prediction," said Dr. Allen. "They never come back with" I do not know "or" I did not find anything "because they are not made for."
She was reluctant to point to individual studies, but said that unsubstantiated findings from the recently published cancer data learning badysis were a good example.
"There are cases where discoveries are not reproducible," said Dr. Allen. "The clusters discovered in one study are completely different from those found in another. Why? Today, most machine learning techniques always say, "I found a group". Sometimes it would be much more useful to say, "I think some of them are really grouped, but I'm not sure about these others." "
advisable
Once machine learning identifies a particular link between patient genes and a characteristic of their disease, human researchers can then develop a scientific rationalization of discovery. But that does not necessarily mean that it is correct.
"It's always possible to build a story to show why a particular group of genes is clustered," said Dr. Allen.
Computer scientists are just beginning to understand the problem, which was threatening to lead medical researchers on false leads and wasting resources trying to confirm results that could not be replicated.
Dr. Allen and his colleagues are trying to improve statistical techniques and machine learning technology so that artificial intelligence can criticize its own data badysis and indicate the likelihood that a given finding will be authentic rather than random .
"One of the ideas is deliberately to disrupt the data, to find out if the results survive this disruption," she said.
Source link
