So far there have been a lot of good suggestions re training for high altitude turbulence, stalling, and even spin recovery. However, as many have pointed out, the first priority is to avoid the extreme weather. There is a long history of jet fighters being lost in large CBs because they lacked any weather radar and yet they were stressed well above airliners' G limits and had much higher control response.
We can't properly simulate really severe turbulence, so the training - as for ditching - has to be rather arbitrary. There is a lot more we can do in regard to weather avoidance at reasonable cost/risk benefit by integrating weather data sources and I hope at least that this tragedy will spur greater efforts in that direction.
It has also become obvious that current automatics in extreme turbulence and icing may not cope adequately and that pilots are easily overstressed by violent movement and temporary or spurious warnings. Hopefully, this accident will provoke better algorithms in preference to relying more on manual takeovers. It's very hard in severe turbulence, blinded by lightning, to take over efficiently with your head and arms flailing around.
Having said that, each pilot must still know the limitations of their auto systems and be prepared for manual takeover. I'll bet on the technology in the long run. Self driving cars have, to my knowledge, not had an accident yet and had completed over 3 million miles last time I checked.