We're, I don't know how many years away from being able to build HAL: we haven't really got any kind of a clue as to how the thinking meat does it, and without some better ideas, well, we're stuffed (as are Google: see the '200 year' comment).
I mean, even the structural-ambiguity of something like the sign I saw the other day ('Do not smoke or spin coins over the table') is so difficult to resolve (here's just a few possibilities) ...
Do not [[smoke] or [spin coins [over the table]]]
don't smoke anywhere, don't spin coins over the table
Do not [[smoke] or [spin coins] [over the table]]
this is the reading they want you to get
[[Do not smoke] or [spin coins [over the table]]]
this is another reading, where the sign is asking you either to not smoke in general or to spin coins over the table
[[Do not smoke] or [spin coins] [over the table]]
unlike the previous one, you can smoke anywhere else but above the table now. But you're only be asked either to not smoke or to spin coins. You're not being asked to not spin coins
Do not [[[smoke or spin] coins] over the table]
you can't smoke coins
Do not [[smoke or spin] [coins [over the table]]]
imagine there are some coins which generally hover over the table. You mustn't smoke or spin those, but you can smoke or spin other coins till your heart's content
How do you get a machine to pick the right interpretation? Welcome to the world of the AI researcher (whose head hurts)!