Isotherms of adsorption of CO2 and Ar are simulated by the grand canonical Monte Carlo on four model surfaces of amorphous silica. The surfaces designated A through D differ progressively in their degree of annealing, A being an unannealed, nonequilibrium surface and D being the most extensively annealed. The gas-gas interaction potentials for both gases were taken from the literature and the gas-solid interactions were modeled by applying Lorentz-Berthelot combining rules to the gas-gas potentials plus the TTAM representation of the atom-atom interactions in the solid. The simulated isotherms of Ar on surfaces A and D are close to each other and to the experimental isotherm for nonporous silica. In contrast, the simulated isotherms and isosteric heats of adsorption of CO2 on these surfaces differ considerably from each other. This leads to the conclusion that argon adsorption is not sensitive to the changes in surface structure that occur during annealing, but CO2 is. Since these gases differ considerably in their polarity, these results indicate that the observed differences in adsorption behavior are due in large part to the annealing-induced changes in the electrostatic part of the CO2-SiO2 interaction. The isotherm of CO2 on D was made to be very close to an experimental isotherm on dehydroxylated nonporous silica by diminishing the electrostatic part of the CO2-SiO2 interaction by 30% from that in the original model of the potential. Isotherms of adsorption of CO2 on multicomponent glass fibers measured at 194.5 K are reported here. The sub-monolayer experimental isotherms on glass lie higher than an isotherm on hydroxylated silica which in turn lies higher than that for dehydroxylated silica. The sensitivity of the physical adsorption of CO2 to the chemical nature and the structure of the SiO2 surface indicates that comparisons of experiment with simulations of the isotherms of polar or quadrupolar molecules like CO2 on such model surfaces can be a useful probe of surface structure of silica and silicate glasses.