Of course there is the even rarer subset: the clue that remains ambiguous even with every crossing letter correctly filled. A setter is very much an author, and your comments have convinced me more than ever that the relationship with the solver is what keeps us coming back.Īs for accidentally ambiguous clues, they are I think rare beasts.
Despite the alarm expressed by one of the setters (none of whom are implicated in the exercise any dudgeon must be directed solely at me), the phenomenon of AI clues that actually work not only seems a long way off, it’s something that holds no appeal to me beyond novelty. I truly appreciate the herculean capacity among you lot to create sense out of nonsense, and to a lesser extent I’m also impressed with the AI’s flair for creating plausible and intriguing surface readings. I had considered removing the letter counts, but had not anticipated that Guardian solvers might be assiduous enough to discover answers that worked if, say, a “(6)” became a “(5)”. Despite the nonveiled threats from some among you, I shall be brave and disclose that the clues were created by feeding a few puzzles’ worth of material from each setter into an artificial intelligence engine and asking it to produce more of the same. As Jacob Busby surmised, the clues were not written by the setters whose names accompanied each set. As many of you suspected, there was something fishy about the request. The best way of finding out, it seemed to me, was to ask the finest minds in solving to look at some clues in isolation. How often does a clue have these properties? It’s hard to say because once the first entry has been written in a grid, the person testing or solving the puzzle tends to use crossing letters for the remaining clues.
Same clues, each one simultaneously offering two different answers.
0 kommentar(er)
