blakestacey @ blakestacey @awful.systems

Posts

50

Comments

1,320

Joined

2 yr. ago

Moderating

---

CHOTINER: Mr Yudkowsky, it says here 20% of your research budget went to “harry potter”. Care to explain?

https://bsky.app/profile/jpeg40k.bsky.social/post/3lzrvedbfe22o

I am Jack's bone-weary sigh.

Damn, these days you can't even post a recipe without a dath ilani guy getting mad in the comments

There are probably some downright horrifying things in Yud's dath-ilangations, but I bounce off the format every time I try to read it, so I won't be the one to find them. I am OK with this.

other fanfic writers in rationalist-adjacent space have complained about his amateurish attempts at conlanging

that feeling when the chess club shoves you into a locker

Bethany Brookshire:

Some people think ChatGPT has a place writing things like news briefs, stuff written to a specific style and tone and, y'know, kinda boring.

So Science did a study.

ChatGPT failed.

Why? It got stuff wrong. "Also, extensive editing for hyperbole was needed." https://www.science.org/content/blog-post/can-chatgpt-help-science-writers

glowfic: it's like a forum, but worse(tm)

"I have a particular set of skills. Mostly 20,000-word blog posts."

If you use physical force to stop me however, I will make it a priority to ensure you regret doing this when you are on your deathbed. You have probably never met an enemy as intelligent, creative and willing to play the decade-long game as I am.

"When you were partying, I studied the blade."

I have added a paragraph about this to the review, since it seems important enough.

translator's note: IABIED means "plan"

And yet people are willing to call him an "AI theorist" or an "AI researcher". That's like calling me the theologian who finally cracked the Problem of Evil with my Hellraiser fics.

It's worse than that: Aaronson doesn't even give a correct description of the Conway--Kochen free will theorem. He says it uses "the standard Bell experiment", but it actually uses a pair of twinned spin-1 particles (maximally entangled qutrits) and a set of measurements given by Asher Peres' 33-ray KS set. Aaronson's review of Wolfram's book doesn't use KS, either, only Bell's theorem phrased as an XOR game. It's like he didn't even read the Conway--Kochen paper, or else he didn't care to explain what was actually in it.

What's the best way for Effective Altruists to allocate grant money? Purchasing malaria nets? Direct giving? No, of course not. It's buying $28k worth of copies of HPMOR (archive)

If I had to guess, I'd say it's One True Nerd Opinion-ism.

"Conspiratorial" implies intent, and I didn't get the sense that 2007--2013 Scott Aaronson was saying that. The tone is more that orthodox physicists are incompetent at explaining things, or indifferent to the need to explain things, or unenlightened to the glorious simplifying power of computerological thinking.

Yeah, doing full-blown quantum physics with the usual mathematical formalism really does require complex numbers, but I don't know of any derivations of that which appeal to computation, for reasons along the lines you indicate.

(It actually all started with Fourier series. Back in 1919 or so, Bohr started speculating that transition rates between atomic energy levels depend on the coefficients in a Fourier expansion. This led, through confusing intermediate steps, to Born's "square the absolute value of a complex number to get a probability" rule in 1927.)

Math competitions need to start assigning problems that require counting the letters in fruit names.

(sees YouTube video)

I ain't [watchin] all that

I'm happy for u tho

Or sorry that happened

This is outside my own department, but I think there's a problem with Aaronson's treatment of Gödel's incompleteness theorems. He says that Gödel's first incompleteness theorem follows directly from Turing's proof that the halting problem is undecidable. This doesn't quite work, as I understand it. The result conventionally known as Gödel's theorem is stronger than what you can get from the undecidability of the halting problem. In other words, the result that the Turing machines get you depends upon a more demanding precondition than "consistency", and so it is somewhat less impressive than what was desired. My best stab at a semi-intuitive explanation would be in the vein of, "When you're discussing the consistency of mathematics itself, you have to be double-special-careful that ideas like the number of steps a Turing machine takes really do make sense."

The historical problem is that Turing himself did not prove the undecidability of the halting problem. He wasn't even focused on halting. His main concern was computing real numbers, where naturally a successful description of a number could be a machine that doesn't stop. The "halting state" as we know and love it today was due to Emil Post.

Moreover, this is one of the passages where Aaronson seems to be offering the one and only true Nerd Opinion. He is dismissive of any way to understand Gödel's theorems apart from the story he offers, to the extent that a person who had only read Aaronson would be befuddled by anyone who used Gödel numbering after 1936.