THE WINOGRAD Schema Challenge is a competition intended to reward technologists who can build a system that understands the kind of ambiguous sentences humans come out with all the time, but which are simple for other humans to understand.
Get it right 90 per cent of the time and $25,000 is up for grabs. And with things like Apple’s Siri, Microsoft’s Cortana and Google Assistant, the Challenge must surely be as good as obsolete by now.
The best two entrants at the event this week achieved correct scores only 48 per cent of the time, little better than randomly guessing the meaning of the sentences they were supposed to crack.
This is despite a decade of advances in the field of AI, which has barely shifted since the late 1950s, according to some.