this post was submitted on 12 Mar 2024
40 points (100.0% liked)

TechTakes

1402 readers
162 users here now

Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.

This is not debate club. Unless it’s amusing debate.

For actually-good tech, you want our NotAwfulTech community

founded 1 year ago
MODERATORS
 

As suggested at this thread to general "yeah sounds cool". Let's see if this goes anywhere.

Original inspiration:

The post Xitter web has spawned soo many “esoteric” right wing freaks, but there’s no appropriate sneer-space for them. I’m talking redscare-ish, reality challenged “culture critics” who write about everything but understand nothing. I’m talking about reply-guys who make the same 6 tweets about the same 3 subjects. They’re inescapable at this point, yet I don’t see them mocked (as much as they should be)

If your sneer seems higher quality than you thought, feel free to make it a post, there's no quota here

you are viewing a single comment's thread
view the rest of the comments
[–] [email protected] 10 points 8 months ago (5 children)

it’s the actual fucking worst when the topics you’re researching get popular in TESCREAL circles, because all of the accessible sources past that point have a chance of being cult nonsense that wastes your time

I’ve been designing some hardware that speaks lambda calculus as a hobby project, and it’s frustrating when a lot of the research I’m reading for this is either thinly-veiled cult shit, a grift for grant dollars, or (most often) both. I’ve had to develop a mental filter to stop wasting my time on nonsensical sources:

  • do they make weird claims about Kolmogorov complexity? if so, they’ve been ingesting Ilya’s nonsense about LLMs being Kolmogorov complexity reducers and they’re trying to use a low Kolmogorov complexity lambda calculus representation to implement their machine god. discard this source.
  • do they cite a bunch of AI researchers, either modern or pre-winter? lambda calculus, lisp, and functional programming in general have a long history of being treated as the magic that’ll enable the machine god by AI researchers, and this is the exact low quality shit research that led to the AI winter in the first place. discard this source.
  • at any point do they casually claim that the Church-Turing correspondence has been disproven or that a lambda calculus machine is superturing? throw that crank shit in the trash where it belongs.

I think the worst part is having to emphasize that I’m not with these cult assholes when I occasionally talk about my hobby work — I’m not in it to make the revolutionary machine that’ll destroy the Turing orthodoxy or implement anyone’s machine god. what I’m making most likely won’t even be efficient for basic algorithms. the reason why I’m drawn to this work is because it’s fun to implement a machine whose language is a representation of pure math (that can easily be built up into an ML-like assembly language with not much tooling), and I really like how that representation lends itself to an HDL implementation.

[–] [email protected] 5 points 8 months ago (1 children)

Oh boy, I have thoughts about Kolmogorov complexity. I might actually write a section in my textbook-in-progress to explain why it can't do what LessWrongers want it to.

A silly thought I had the other day: If you allow your Universal Turing Machine to have enough states, you could totally set it up so that if the first symbol it reads is "0", it outputs the full text of The Master and Margarita in UNICODE, whereas if it reads "1", it goes on to read the tuples specifying another TM and operates as usual. More generally, you could take any 2^N - 1 arbitrarily long strings, assign each one an N-bit abbreviation, and have the UTM spit out the string with the given abbreviation if the first N bits on the tape are not all zeros.

[–] [email protected] 4 points 8 months ago

You could use the recent-ish Junferno video about Turing machines to demonstrate that point as well

[–] [email protected] 3 points 8 months ago* (last edited 8 months ago) (1 children)

doing work that's not trying to free us from the tyranny of century-old mathematical formulations? how dare you! burn the witch!

(/s, of course! also your hardware calculus project sounds like a nicer time than my batshit idea (I want to make a fluidic processor.. someday..))

[–] [email protected] 4 points 8 months ago (1 children)

I want to make a fluidic processor… someday…

fuck yeah! this sounds like the kind of thing that’d be incredibly beautiful if done on the macro scale (if that’s possible) — I love computational art projects that clearly show their mechanism of action. it’s unfortunate that a majority of hardware designers have a “what’s the point of this, it’s not generating value for shareholders” attitude, because that’s the point! I will make a uniquely beautiful computing machine and it won’t have any residual value any capitalist assholes can extract other than the beauty!

if I ever finish this thing, I should make a coprocessor that can trace its closure lists live as it reduces lambda calculus terms and render them as fractal art to a screen. I think that’d be fun to watch.

[–] [email protected] 3 points 8 months ago

Yep. I love beautiful machines with beautiful action in the same way.

One of my favourites I’ve seen was a clock with a tilt table, switchback running tracks running widthwise across that table, and switches by the track ends. A small ball would run across the track for 60s until it hits the switch, which would cause a lever system to flip the orientation of the tilt table (starting the ball movement the other way).

Saw it in the one London collection of typically-stolen antiquities, I don’t recall the origin of it.

For the processor: yep, something larger is the intent, but I think I’d have to start with a model scale first just to suss out some practical problems. And then when scaling it, other problems. God knows if I’d want to make this “you can walk in it” scale, but I’ll see 😅

[–] [email protected] 3 points 5 months ago

lisp (...) [has] a long history of being treated as the magic that’ll enable the machine god

Nonsense, we all know the robot god will be hacked together in Perl.