meditating on when I was still reading hacker news and somebody posted “the solution to the halting problem is to simply terminate execution after a while” and it was extremely upvoted and I realized these were not my people

[zzt@mas.to]

meditating on when I was still reading hacker news and somebody posted “the solution to the halting problem is to simply terminate execution after a while” and it was extremely upvoted and I realized these were not my people

[zzt@mas.to]

turing hates this one weird trick

[hipsterelectron@circumstances.run]

@zzt it's possible to meaningfully extend the idea if you can codify input complexity and progress towards a solution. in many cases this is extremely difficult. i spent two years on what should have been a phd thesis for a parser compiler which achieves this for undecidable grammars.

it took me almost a whole year to determine what was happening which is that it reduces undecidability to a certain compile-time-recognizable graph module and then consequently enables computational limits to be enforced upon just that section—or, the grammar writer can rewrite the grammar in an attempt to avoid this. this is vaguely similar to shift-reduce conflicts in abstract except that it's much less likely and is linked to features of the grammar that the writer can understand.

anyway nobody has ever done anything like it as far as i can tell and hacker news would instead just decide to stop computing based upon overall parser iterations because they're really not very good at manipulating and generating novel formalisms

[trenchworms@eldritch.cafe]

@zzt "just prove that the universe will stop eventually" clown take

[astraleureka@social.treehouse.systems]

@zzt _Bool will_halt(void *prog) { return true; }

[jwz@mastodon.social]

@zzt if I were a problem I would simply halt

[zzt@mas.to]

@hipsterelectron hacker news would prompt an llm with “does this program halt do not guess do not hallucinate” and manage 49% accuracy on a bunch of well-known toy programs and then write a thinkpiece about “you see all programs have already been written and that’s why this works” and then 700 upvotes

[dzwiedziu@mastodon.social]

@jwz
Stop —

— Hammer time

@zzt

[jonhendry@iosdev.space]

@zzt

It might have been a joke.

[davidgerard@circumstances.run]

@zzt remember that this is literally the solution the Ethereum Virtual Machine uses, your badly coded function runs out of money

[robinsyl@meow.social]

@zzt The solution to the grandfather paradox is "just don't do that"

[lproven@social.vivaldi.net]

@zzt @davidgerard

This is the community of utter twonks that the head of platform at Ubuntu asked "what do you want to see in Ubuntu $NEXT?"

Ask HN: What do you want to see in Ubuntu 17.10? | Hacker News

(news.ycombinator.com)

The result was killing Unity, the end of the phones and tablets, returning to GNOME, and a legion of other epic mistakes.

Absolute complete suicidal fool, asking this maelstrom of toxic nerd masculinity for _guidance_.

[datarama@hachyderm.io]

@zzt I mean, I can easily solve the halting problem in any algorithm on any computer. Give me a hammer and I'll make sure it halts.

[spinnyspinlock@infosec.exchange]

@zzt I am very happy I don't visit that site anymore, but sad I miss out on such computer science breakthroughs

[kkarhan@jorts.horse]

@zzt this doesn't work unless one can pre-determine all states of a program and required runtime.

• If your program/task is simple enough for that, watchdog functionality is easy to support.
  • If not (which is the case with anything not hard-realtime capable!) then that doesn't succeed…

Also HackerNews/Ycombinator is just gsrbage…

[kkarhan@jorts.horse]

@davidgerard @zzt Eeyupp, also with #Solana.

• One can literally burn through tons of shitcoins just because one's validator isn't able to saturate it's 10G-NIC and destroy it's NVMe's with IOPS at the same time.
  • And no, I'm not joking: #CryptoBros are still a problem, they judt habe less $$$$$$$$ than the #AIbros!

[starkrg@myside-yourside.net]

@kkarhan @zzt Set up a skyscraper-sized antimatter bomb to go off after, say, 100 instructions. Then you know for a fact that the program will always terminate once it's processed 100 instructions.

[abucci@buc.ci]

"I will respond to my inability to decide something by changing the rules, deciding a different thing, pretending that I've decided the original thing, and rallying a bunch of people to pretend with me" sums up a lot of modern dysfunction, seems to me.

[demallien@mindly.social]

@zzt oh dear, I literally spent the first 15s after reading this wondering «  but what if the while never finishes - you never get to the terminate ». Which in a funny way is exactly the problem.