The range of warming predicted for the globally averaged surface temperature as a result of a doubling of CO2 due to human activities has been around 1.5-4.5 degree Celsius for over 20 years. As we learn more about the climate system, the corresponding models become more detailed and realistic, but the range of predicted warming has not decreased substantially. The aim of the workshop is to discuss explicit ideas on reducing uncertainty. Such ideas can be classified into two groups, which we might call reductive and holistic. The first consists of improving the parameterization of processes that occur on a scale too small to be modeled explicitly, such as ocean mixing, clouds, etc. The second uses observations (past climate variability on all time scales, the observed response of climate to recent volcanic eruptions such as Pinatubo, the warming over the 20th century, etc) to tell us something about climate sensitivity. The workshop will combine individuals involved in the effort to minimize climate prediction uncertainty using both approaches, in the hope of stimulating a discussion between these two communities, and to educate a group of advanced graduate and postdoctoral students to be aware of both approaches, their strengths and their weaknesses. Specific hypotheses for the lack of progress in reducing uncertainty will be discussed, including arguments on the structural stability of climate models with respect to parameterized processes, as well as approaches which might be taken to address this problem in the near future. The format will consist of a few detailed lectures a day, allowing for a large portion of time dedicated to informal discussions among the lecturers themselves, as well as between the students and lecturers.