Using ZFC, Robinson can take a standard model of the reals and construct a nonstandard model, and a function from the standard model to the nonstandard model. Every possible statement of first order logic that can be made about the standard model is true in the nonstandard model. Likewise every possible statement of first order logic that can be made in the nonstandard model only involving things that came from the standard model are also true of the standard model. This is his "transfer principle". However the nonstandard model comes with things like infinitesmals, and infinite numbers. So in the nonstandard world we can really do things like define a derivative as dy/dx, or an integral as an infinite sum. Then recover the classical definitions through the transfer principle.
See https://terrytao.wordpress.com/2007/06/25/ultrafilters-nonst... for an explanation of how this actually works in some detail. But if you get your head around it, you'll see exactly why first order logic cannot possibly capture the idea of being a standard model. Because, in first order logic, the nonstandard model really is indistinguishable. And so finite cannot be expressed in first order logic.
Using ZFC, Robinson can take a standard model of the reals and construct a nonstandard model, and a function from the standard model to the nonstandard model. Every possible statement of first order logic that can be made about the standard model is true in the nonstandard model. Likewise every possible statement of first order logic that can be made in the nonstandard model only involving things that came from the standard model are also true of the standard model. This is his "transfer principle". However the nonstandard model comes with things like infinitesmals, and infinite numbers. So in the nonstandard world we can really do things like define a derivative as dy/dx, or an integral as an infinite sum. Then recover the classical definitions through the transfer principle.
See https://terrytao.wordpress.com/2007/06/25/ultrafilters-nonst... for an explanation of how this actually works in some detail. But if you get your head around it, you'll see exactly why first order logic cannot possibly capture the idea of being a standard model. Because, in first order logic, the nonstandard model really is indistinguishable. And so finite cannot be expressed in first order logic.