Followup — The AI Researcher & Harry Potter
This continues from yesterday’s post, and will make no sense without it.
I was wrong. Completely wrong. There, no words minced. One commenter said I had fallen for a joke, which may be true. It wouldn’t be the first time. The intended solution involved talking only as a stalling tactic and no agreement. (I speak only of the solution for getting out of the death trap … the ending still has a ways to go).
I do take some consolation from the Author’s comments: (note, link has spoilers)
On a larger scale, the verdict is in: your collective literary intelligence has exceeded mine. There were at least half a dozen brilliant ideas I’d never imagined…..To be even remotely solvable to the individual reader, the story needed to use the heavily foreshadowed solution… Most readers not connected to the Internet community did not solve the dilemma, and their initial responses were often “AAAHHHH IMPOSSIBLE”. It wouldn’t be fair to those individuals readers to hit them with your more awesome and less predictable outcome – but your stuff was indeed cooler…You clearly could have done this without my having tried to deliberately set up a solution in the text, and you still would have solved it. But I didn’t know that back when I was planning the whole story,…
So … Not the AI problem; still an experiment on collective intelligence. Yudkowsky also admitted (later in the thread) that he’d never written the shorter/sadder ending, because he was confident the group would figure it out his intended solution, consider it too obvious then spin out ludicrously more complex solutions and awesome solutions (working and not). I saw one comment claiming the correct solution was posted several hours prior to the exam.
One brilliant posted solution (spoilers, obviously) seems reasonable and involves the sorting hat, a time turner, the cloak of invisibility plus an unbreakable vow, Mendelian inheritance, solving the p=np problem, and an equally important paper cut (all of which were discussed in prior chapters). So now I don’t feel so bad about my first ridiculously complex solution, either.
I’ll try to learn from my mistake. I probably won’t, though.
Update — According to this thread, the proposed solutions had a final word count of just under 775k.
“I have the distinct feeling that I ought to be doing something about this,” sighed Professor McGonagall.
“Ignore it, it’ll go away. Ooh, shiny!” — McGonagall and HJPEV, Chapter 6 “The Planning Fallacy“