Great article. Do you think an MoE approach could foster an open(ish) ecosystem in which researchers in academia/industry can contribute and share individual models? Downstream aggregators could pick and choose from this marketplace based on their specific requirements (drug discovery, diagnostics, predicting protein interactions, etc.). Lots of legal, ethics, financial issues to work out in addition to technical ones.
My two cents, I think it might be tricky to pull this off. A lot depends on assay/experimental conditions. Mixing different models can yield inconsistent results. Not to mention IP issues which will be major of course
Great article. Do you think an MoE approach could foster an open(ish) ecosystem in which researchers in academia/industry can contribute and share individual models? Downstream aggregators could pick and choose from this marketplace based on their specific requirements (drug discovery, diagnostics, predicting protein interactions, etc.). Lots of legal, ethics, financial issues to work out in addition to technical ones.
My two cents, I think it might be tricky to pull this off. A lot depends on assay/experimental conditions. Mixing different models can yield inconsistent results. Not to mention IP issues which will be major of course