EDIT
Variations of atmospheric CO2 occurring as a climate feedback on the time scale of the ice ages (Figure 3) can be ~100 ppm in 5000 years, or 0.02 ppm/year. This atmospheric change is due to a shifting of carbon among the atmosphere, ocean, soil and biosphere compartments within the surface carbon pool, a warmer climate driving more CO2 into the air. This natural glacial-interglacial variation of atmospheric CO2 is quite rapid in comparison with the geologic cycling of carbon between the Earth's crust and the surface carbon pool, which amounts to ~10**(-4) ppm/year of CO2, as discussed above.
These natural rates of atmospheric CO2 change must be compared with the human-caused growth of atmospheric CO2, which is now ~2 ppm/year (see below). Humans, indeed, are now in control of long-lived atmospheric GHGs. As a result it is important to investigate climate sensitivity for the case in which GHGs are specified as the forcing. The Charney climate sensitivity applies to this case under the assumption that slow feedbacks such as ice sheet area, vegetation distribution, and climate-induced GHG changes are not allowed to operate.
As a complement to the Charney climate sensitivity, let us derive the climate sensitivity that applies if these slow feedbacks are allowed to operate: we call this the "long-term" climate sensitivity. We can obtain this "long-term" climate sensitivity from paleoclimate data by finding the scale factor that causes the GHG forcing to match the paleoclimate temperature change as accurately as possible. Figure 4 shows that multiplying the climate forcing due to long-lived GHGs (CO2 + CH4 + N2O) by 3.02°C per W/m2 yields remarkably good agreement with Antarctic temperature. Given that glacial-interglacial global temperature change is about half of Antarctic temperature change, this implies a "long-term" climate sensitivity of ~1.5 W/m2 or about 6°C for doubled CO2.
Which climate sensitivity is more relevant to humanity: the Charney 3°C for doubled CO2 or the "long-term' 6°C for doubled CO2? Both. The net human-made climate forcing, including negative forcing by tropospheric aerosols, has been substantially positive only for the past three decades. On that time scale the Charney sensitivity is a good approximation, as little contribution from slow feedbacks would be expected. Thus climate models with 3°C sensitivity for doubled CO2, incorporating only the fast feedbacks, are able to achieve good agreement with observed warming of the past century. We suggest, however, that these models provide only a lower limit on the expected warming on century time scales due to the assumed forcings. The real world will be aiming on the longer run at a warming corresponding to the higher climate sensitivity.
EDIT
http://gristmill.grist.org/story/2007/10/1/131321/083Link to full Hansen article:
http://www.columbia.edu/%7Ejeh1/East-West_070925.pdf