A recent paper in Nature by Mathews et al proposes a new mathematical construct for considering the increases in temperature following carbon dioxide emissions. In additional to the typical temperature response depending on increased concentrations of atmospheric CO2, their model takes into account the changes in atmospheric CO2 concentration following changes in CO2 emissions. Essentially they have added a feedback parameter to the standard account.
In their model the total amount necessary to raise global temperatures 2 °C above pre-industrial levels is a mean of 1.4 trillion tons of CO2 emitted, with a 95% confidence interval of 1.0 to 1.9 trillion tons. Currently the world has emitted 0.5 trillion tons of CO2, so in order to stay below the 2 °C level we would need to, on average, limit emissions to 0.8 more trillion tons. Note that zero net future anthropogenic CO2 emissions would be necessary to fully stabilize global mean temperature. That's because cutting emissions would not quickly reduce atmospheric concentration; atmospheric carbon dioxide does not act like your typical pollutant.
Robin Hanson favors geoengineering, which sounds good to me as a temporary fix. Per Eric Drexler's recommendation, we could use thermodynamically efficient molecular pumps in the long run to reverse net atmospheric concentrations. Neither of these are politically favorable because: 1) Most people do not know about them, and 2) No existing set of corporations stand to directly benefit from their development and implementation. Still no villians, but yes unfortunate incentives.