Zerouali--your post shows a few misunderstandings.
First, that night was clear and cold. I flew out of PANC early the next morning I can tell you that, unless there's fog, when it's very clear the visibility is very good: more than 10sm, which would be the max reported by the ASOS (ATIS).
Second, during low workload it's common for the controller to simultaneously work both ground and tower. They often will clear an aircraft for takeoff while that aircraft is still on the taxi way. That's exactly what happened in this case except the clearance was received while they were taxiing on the *parallel* taxiway. The crew acknowledged the clearance, made a 90 degree turn on to the taxiway that was supposed to take them to RWY 32 and then began the takeoff run.
From the tower controllers point of view the 90 degree turn was expected. The accleration towards Cook Inlet was not!
I'm sure all of this is archived somewhere.
It must all boil down to basic airmanship if you ask me. The only problem I can see with ANC ATC *might* be to wait until the aircraft is physically in position on the runway. Of course low-vis ops will pose certain problems with such a procedure but there's always ground radar...
...and I shiver thinking about Milan
Just listen *and* think!