What If We Could Fight Coronavirus with Artificial Intelligence? [pdf]

We absolutely can, on all fronts.

Deep neural networks have emerged as applicable to a wide range of general problems - they are very powerful universal function approximators and excel in noisy data environments.

1. Snooping out trends and correlations, especially nonlinear, non first order interactions that cannot be captures with typical statistics like averages and standard deviations. This could help with predicting mortality, tracking disease, modeling outcomes...

2. Accelerating all manner of modeling, 2D and 3D, e.g. protein folding

3. Classification problems, like image analysis, e.g. diagnosing 2019ncov from lung scans much more quickly and effectively than overburdened staff

Most importantly the technology is here. It is mature. Problems of similar complexity have repeatedly been solved in literature and most of the source code is available for rapid adaptation - and the great thing about neural nets is beyond basic architectural flow, neural nets of correct arrangement for a particular problem need little to no tweaking for other problems within a class. Much in the same way that the same mathematical equations govern phenomena across physics.

Now is our time to mobilize, to make connections and orchestrate solutions. You can run and train most of this stuff on a mid-range GPU with python, not to mention cloud computing is available for big problems.

This is a potential front in the coming civic mobilization against a global pandemic. Even as developers working remotely we can contribute to the effort.