The last sentence says it all.
“My view is we should always choose knowledge over ignorance,” Keutsch said."
They mean go back to square one and get much needed input to climate models.
The entire study is an admission that climate computer modeling is alarmingly simplistic and misleading. The time, talent and money needed to perfect the modeling to actually predict long term global climate usefully is a thousand times larger than learning and attempting to justify playing with climate locally. Climates change and we are 100 years away from predicting that. In some respects that is good news because it probably means that the so called AI singularity is 150 years off at least.