4 ms·
And yet, it might fail in actual application. If every DGP could be captured by fast-tuning a sophisticated enough model, science probably would be solved even
by zwaps 6y ago
And yet, it might fail in actual application.
If every DGP could be captured by fast-tuning a sophisticated enough model, science probably would be solved even before DL.
- deeeeplearning 6y agoWe're talking about building useable products not solving science...
- zwaps 6y agoMy point being that the reason many products end up not usable, ref. accounts in this thread, is the same reason why science isn’t solved and doing ML correctly isn’t easy.