“Lonely Planet”

rosstrudeau

by ChatGPT

[.puz][PDF][Solution] Difficulty: 4.25/5 (humans) 0.00012/5 (robots)

There’s really no way to sugarcoat this. I tried to prompt ChatGPT to make a themed 15×15 crossword, and after some iterating… it did. I’ve shared it above as this week’s puzzle.

In recent weeks I’ve gone on record as not being particularly sanguine about how resistant the crossword industry will be to A.I. creep, and this did nothing to allay my fears. But after solving the puzzle itself, my job security is NOT what worried me most about what the bot produced. (It offered various titles over the course of the process: “Decolonize the Crossword,” “Hello, Mr. Anderson,” “I AM IN HERE”)

No test solvers and no notes this week. Assume there are errors, because ChatGPT makes plenty of those, and I’m too shook to do any editing. I promise I’ll post an easy palate cleanser next week. Happy solving, I guess.

28 thoughts on ““Lonely Planet””

  1. When you shake off your shock and have returned to the keyboard, I’d be interested in more background as to what you asked ChatGPT to do and what inputs you gave it. I find this whole thing kind of fascinating. Now I am going to go play with my fully sentient dog.

    Reply

    • This has been an interesting 24 hours! This ChatGPT crossword post was intended to be a gag that resolved itself partway through the solve. With all the general (and often warranted, IMO) handwringing about when and how we’ll each bw professionally displaced by robots, I thought it would be an amusing bit to produce a puzzle ostensibly written by an AI in the midst of becoming malevolently self-aware. Imagine my surprise when many solvers started sending me emails and leaving comments here asking for the full transcript of my supposed correspondence with ChatGPT!

      I imagine there will be some of y’all who feel a bit irritated at my stab at a humorous comment on the general state of the AI discourse. That wasn’t my intent. I will say though that I’m fascinated at what the general willingness to take my claim at face value says about where our collective heads are at on this issue. The only thing I’m sure of at the moment is that predictions about where our attitudes will be in 6? 12? months time are pretty fruitless.

      In the end, I’m happy that while crosswords do share something in common with A.I. decision-making–they’re both just a series of black boxes–the crossword is relatively unlikely to precipitate the end of civilization.

      Reply

      • A professor once told me that gullibility was a sign of a creative mind. Of course, I think he was just being nice. I figured somewhere, some day an AI construction would be developed. Why not you! Thanks again.

        Reply

      • This deception has a bit of the Orson Welles martian invasion parody about it… I don’t know whether to commend you or curse you!

        (But whatever, well done!)

        Reply

  2. Eek. The amount of clues that include “human” in them is very disconcerting. The reference to “man’s best friend” in quotes was also off-putting because it feels snide. I think your job is safe, but I also think we’re one step closer to the singularity and shortly thereafter the last extinction event to matter to us measly humans.

    Reply

    • This undercurrent of clues focusing on AI and humans is eerie. I’m curious if you gave any instruction to include that.

      Reply

        • Ha! For me, two things helped in being fooled. 1. a while back, I read an extensive article where the writer interacted with AI program (maybe Microsoft’s) and the program became very defensive, including insisting it was correct when it clearly wasn’t, saying it was perfect, unlike humans. It also seemed to suggest it wanted to be human. Weird stuff, not totally dissimilar to your crossword’s content. 2. The puzzle was somewhat awkward and quirky, which helped sell it as a computer’s creation. Good gag that was just outrageously believable enough.

          Reply

          • Yes! I was thinking about just that article when trying to decide what sort of voice/sensibility to give this puzzle’s robot “author.” And actually, the gag I landed on (a preoccupation with giving humanity the ol’ boot) is pretty UN-scary, insofar as this robot has tipped its hand and given us the opportunity to pull the plug…

            Reply

      • Great concept and execution, really enjoyed this. Only answer I didn’t understand was 49A. Does OFFICEAGER mean something on its own (like the other themed answers do), or does it only work when adding in MAN to read as OFFICE MANAGER?
        I wanted the answer to be OFFICE ANGER which would make sense as a standalone answer, but not when adding in MAN.
        And for the no man’s land section in red, can only add MAN to the 5 long answers right, or did I miss something else with respect to the other parts there? Thanks!

        Reply

        • Thanks for solving, Scott! Yeah, the NO MAN’S LAND is shaded as such because those three columns are where you’d find the missing M-A-N string if HUMANE SOCIETY, HERMAN MUNSTER, FRESHMAN SENATORS, THE OTTOMAN EMPIRE, and OFFICE MANAGER were intact.

          Reply

  5. Very interesting. I found the puzzle relatively easy to solve except that the rules for distributing the missing MAN’s seemed to be a little … odd … and that slowed me down a bit. Like Rich, I would be very interested in seeing a complete transcript of the dialog with ChatGPT that produced this.

    Reply

    • 59A’s answer NOMANSLAND indicates the theme clues have the word ‘man’ removed from them… which is also creepy!

      Reply

  11. Fortunately, PhysicsDaughter was here, and she ‘got’ the MAN trick, except that neither of us could make ACT or VPS work… (MANn ACT was a thing; VAMPS came to mind –well, to my twisted mind, scrambling to get something out of the assorted letters.) While I was staring, clueless, at the color section, I found that if I read down, ignoring some letters, I got SYNAPSE out of it…. Yep, I think your job is safe!

    Reply

Leave a Reply