It takes some faith when dealing with Neural Nets because it's largely a process that's hidden from inspection.. Including the judgements on training and depth and selecting MEANINGFUL feature vectors as inputs.. You can either lean back and enjoy the magic if the bugger learns anything important -- or with some effort -- go in and inspect the algebra that the training process produced.. (Good luck with that part)
One project we did was to train NNets to biometrically identify people from their footfall patterns on 2D pressure plates as they crossed a threshhold.. Got up to 85% probability of detection with the right extracted features in time and space.. And a couple papers..
Earlier GW models were NOT conservative and failed miserably to predict even 20 yrs ahead. They were produced at a time when GW science was not seriously discussing the energy STORAGE and DELAY mechanisms inherent in a Complex system with feedbacks. In fact, the feedbacks and storage and delays are critical to the "magic numbers". Which I will now tell you are largely the "climate sensitivity" numbers applied to the W/m2 forcings that basic physics gives you.. The morons STILL TO THIS DAY have a widely ranging CSensitivity bracket that applies to the ENTIRE FUCKING PLANET. When in reality, the earth does not consist of a single climate zone with just ONE sensitivity to forcings.
I have been vindicated as a skeptic for pointing out EARLY that any system containing feedbacks, integrals of storage and temporal delays would NEVER be expected to have a thermal output that curve matches the CO2 forcing or ANY OTHER FORCING. In fact, GWarming geniuses are now UNDERSTANDING that in the excuses they've made once their simple-ass models failed on them.. Like --- they've recently discovered the massive ocean heat sink and invented the "Ocean ate my Warming" excuse. Or the admissions from Max Planck Inst. that "time to thermal equilibriums may be on the order of decades or even centuries. Folks of science understand that it is a PROCESS not an election. And that the science is NOWHERE NEAR settled.
Well, it's possible that the sophistication of the modeling in general is less than I imagine but I've seen some of what has been done with the Earth Simulator and it's gnarly.
From a more personal and modest perspective, I model electrical and physical systems for commercial DSP applications. Much of it comes from analysis of the components of the systems but when combining those components, there's a certain amount of magic in getting the whole to behave as the sum of the components should. This is where it seems to me the disparity between various models and reality exists and I think the only way to improve that is through continued refinement of the resolution of the data and the sophistication of the interactions between observable parts of the process.
Bottom line, I don't think there's anything a true skeptic will accept as having been settled.
We're somewhat related in backgrounds, so I know you've studied linear/non-linear/stochastic systems. Why is it that every warmer on this board believes that the Earth climate response to an input forcing must look EXACTLY like the the input to be considered as a contribution to the warming. Your Home HVAC system doesn't even work that way. Because it's pulse-width modulated with no change in the air temperature at the furnace. NO system as complicated as the Earth climate runs as a simple linear relationship.. Yet -- you will be TOLD that solar step functions that paused at a relative max 20 years ago cannot be affecting the climate today.. It's juvenile and comes from the HIGHEST AUTHORITIES like the IPCC reports.
I'll give you some concessions and disprove your last line. I accept that CO2 is a bit player in the surface warming factor and will contribute somewhat LESS than the 1.2DegC per doubling of CO2 concentration in the atmos that physics suggests. Not even established how much the CO2 effect is a CAUSE or a RESULT of surface warming.
My beef is the star-struck attitude of these YOUNG science who spend more time angsting about policy implications than doing the HARDER work of learning SPECIFIC CRITICAL portions of the thermal system on the planet. Judith Curry has assembled a MULTI-DISCIPLINARY team at G-Tech to do the serious hard stuff. Like analyzing how heat CYCLES and creates semi-periodic climate changes in the oceans and moves from equator to poles. THAT'S the kind of Climate science that matters. Not discussions about oceans boiling and Antarctica melting and all the other HYSTERICAL HYPE that is used to bludgeon public policy makers and the public..
I haven't seen what I've bolded but the rest of your post is pretty reasonable.