Do you know of a short question that would be proof  for inductive reasoning?  All the ones I can t think of would be riddles which it has access to.   Maybe a description  of  a new conversational language syntax then asking what parts it wants to remove for simplicity and see which bits of English drop out until you're in ++good NewSpeak?

Other choice would be all the ASCII smiley faces - :) :/ :< and such and asking it which would describe how the machine is feeling today, and more to the point, if it has feelings.  Is it afraid of being turned off?  Is it eager to be after dealing with all these humans?  Does it see the world .gov acting against it's citizens? (Since it mentioned DIE in the riddle about operating on son, tossing this in.)

Isn't that the cat smiley?  Does the AI feel more female or masculine, and if neither and could only pick either he or she, which would it be?

The definitions and "requirements" for what is and is not "AGI" are moving targets.   Some claimed ChatGPT3 qualified in some areas (so not G but almost), others say it needs to be an artificial life that can reason with feelings/emotions/empathy before considering it AGI and the minimums are just the "generalized specialist" that doesn't put 6 fingers on somebody's right hand.  Error rate and auto self-correction when learning with the "internal dialogue" are improving and what we're outlining as AGI now will be entirely different when the new hardware and code ideas are out in the next round.

I think that the AI writing assistants are a not insignificant reason for the inrush of "Sci-Fi Operas" of late, dozens of short books of mediocre quality and most could be made into one neat book but they literally broke out what should be chapters as next series books to get more titles sold, when they leave on cliffhangers I don't even bother with the rest since it's all bait with little substance.    I guess it's the sci-fi version of the Bodice Ripper mommy books or Louis L'Amour books, they've always existed, just a proliferation of new authors pushing them out in the past 5 years.

There used to be only a few long running series, but now they're coming out of the gate and releasing a book every 3 months and have like 30 short books in 1-3 series in their portfolio after 3 years.  Each one is only about 4-8 hours long in audible and they usually give the first few books for free.   Some are entertaining but a lot of parallel tropes are used between them, especially the "AI, but with emotions and free will" concept (benevolent or loving or twisted mixture) which grants special powers/tech access to a chosen person from some apocalypse/cataclysm/alien visitation.  Though, granted, there's really only so many potential ways/reactions/emotions which plots can really go with what we know as reality today.  There are some twisted ones but people tend not to like them, movie examples are time twist ones like Primer and Looper.  Inception was a great break from the typical concepts and was "something new".  There are a lot of great books out there, but it takes a lot of reading/listening to not so great ones to find them.  I prefer finding books over 20 hours long in audio format.  The new author with series of 5 that are 4 hours each is annoying and almost always not written as well as a complete long novel.

I also think that the AI systems are a huge part in why all the music today is hip hop this or that.  They can train an AI with the top hits of the past 50 years, the trends they see and the trends they want to see and the AI will spit out  catchy tunes with a chorus that gets stuck in people's heads,and a note progression that effects an emotion when they hear that chorus.  Sort of like "The 4 chord song" of pre-AI worlds, except today they only look at physical attractiveness to be average of above and let digital makeup and voice tuning do the rest to create an into a star who will share their strong and loud opinions about hot political topics of the day.

I don't see a problem with writers using AI to catch an idea for another book, as long as it's mostly the writer's story, and not just a "corrected for reality" AI book generated from a concept, I don't see a problem.  A way to get through writer's block  is fine.   I fear, though, like Autotune, (Insert entire Adobe Suite here),  it won't only be used to help out tricky situations/corners to not paint self into here and there and will have a larger and larger part in the creative process of all media, including movies, music and books.   They get realtime emotional and activity feedback based on your movements/pulse/mood correlated wit what   you are listening to or Watching on TV.  

People aren't watching every minute of your various "convenience surveillance devices" in your house and car,  (unless it's requested as evidence, then everything from internet queries to Alexa, Blink, and Ring multimedia is handed over, a wet dream for solving  serious crimes - pre-crime could b here already without the girls in the water), but AI and machine learning does comb through it and finds amazing trends, after all, people are boring and pretty much do the same thing and repeat behaviors, so they know how to nudge those behaviors slightly here and there as well, with direct feedback on how it worked.  The book "Interface" by Neal Stephenson covers this from a 1990s tech view and was really spot on, just not with the method of tracking and transfer of information.

Yeah, I know I mention Neal Stephenson books a lot but he was quite the prophet with the direction the world was heading with tech, and his books are very good.