Move over Dijkstra: New algorithm just rewrote 70 years of computer science • Hacker News Board

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#21

Post by n4r9 » Fri, Oct 03, 2025, 7:47 PM UTC

OutOfHere wrote: post #20

n4r9 wrote: post #19

Earlier quoted context omitted.

Why "evil"? I'm pretty confident they're heavily (if not fully) relying on LLM-generated text. Maybe they're drafting it themselves first and getting an LLM to refine. I found some recent articles by the same author which gave me the same reaction: https://medium.com/@kanishks772/computer-scientists-just-bro... and: https://medium.com/@kanishks772/why-your-next-gps-might-use-... They seem to have a process for grabbi…

It is evil because it is intended to suppress discourse for irrelevant reasons. It is also deceptive because you don't exactly know the level of effort that went into it. It should make no difference it's written by AI or not. One should evaluate and criticize content without regard to who or what has written it. Conversely, just because a human writes something should not and does not make it superior. People like y…

I'm not trying to suppress anything, but I want people to be aware. Conversely, I would gladly point out that an article on mathematics written by Terence Tao is likely to be high quality and insightful. Such guides are helpful in a world where not everyone has the time or energy to fully get to grips with the minutiae.

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#22

Post by OutOfHere » Fri, Oct 03, 2025, 8:05 PM UTC

n4r9 wrote: post #21

OutOfHere wrote: post #20

Earlier quoted context omitted.

It is evil because it is intended to suppress discourse for irrelevant reasons. It is also deceptive because you don't exactly know the level of effort that went into it. It should make no difference it's written by AI or not. One should evaluate and criticize content without regard to who or what has written it. Conversely, just because a human writes something should not and does not make it superior. People like y…

I'm not trying to suppress anything, but I want people to be aware. Conversely, I would gladly point out that an article on mathematics written by Terence Tao is likely to be high quality and insightful. Such guides are helpful in a world where not everyone has the time or energy to fully get to grips with the minutiae.

No, what you're doing is unfairly biasing and inciting people against AI. No author should be given a free pass. AI has done a lot more to improve my life than Terence Tao.

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#23

Post by whatamidoingyo » Fri, Oct 03, 2025, 8:21 PM UTC

OutOfHere wrote: post #20

n4r9 wrote: post #19

Earlier quoted context omitted.

Why "evil"? I'm pretty confident they're heavily (if not fully) relying on LLM-generated text. Maybe they're drafting it themselves first and getting an LLM to refine. I found some recent articles by the same author which gave me the same reaction: https://medium.com/@kanishks772/computer-scientists-just-bro... and: https://medium.com/@kanishks772/why-your-next-gps-might-use-... They seem to have a process for grabbi…

It is evil because it is intended to suppress discourse for irrelevant reasons. It is also deceptive because you don't exactly know the level of effort that went into it. It should make no difference it's written by AI or not. One should evaluate and criticize content without regard to who or what has written it. Conversely, just because a human writes something should not and does not make it superior. People like y…

I see absolutely no reason to be this upset at the OP's comment... unless you're the author (erm, publisher?) of the article. Are you?

> It should make no difference it's written by AI or not.

It absolutely should.

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#24

Post by twojacobtwo » Fri, Oct 03, 2025, 9:43 PM UTC

OutOfHere wrote: post #12

n4r9 wrote: post #8

The linked medium post is clearly written by an AI but I think it does a decent job at summarising the results. Glancing through the paper on ArXiv, it feels like they've cleverly combined speed-up techniques from variations of Dijkstra invented over the years. The Thresh X2 [0] algorithm - for example - does away with the priority queue that is the bottleneck in Dijkstra. Instead, it iteratively runs a "label-correc…

Why don't you focus on the actual content instead of attacking its authorship? Your comment is clearly written by an AI.

You need to take a break from the internet for a while. You're getting far too worked up over a passing comment. It clearly wasn't dismissive of the content and, as pointed out, OP immediately followed the comment by talking about the content (I.e. saying it's a decent summary).

Some of us think that anything created by an AI should be labeled as such. Is that an inherently evil belief?

It is a little dumbfounding to compare the seeming emotionality/outrage of your posts with that which spurred it.

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#25

Post by OutOfHere » Sat, Oct 04, 2025, 3:51 AM UTC

whatamidoingyo wrote: post #23

OutOfHere wrote: post #20

Earlier quoted context omitted.

It is evil because it is intended to suppress discourse for irrelevant reasons. It is also deceptive because you don't exactly know the level of effort that went into it. It should make no difference it's written by AI or not. One should evaluate and criticize content without regard to who or what has written it. Conversely, just because a human writes something should not and does not make it superior. People like y…

I see absolutely no reason to be this upset at the OP's comment... unless you're the author (erm, publisher?) of the article. Are you? > It should make no difference it's written by AI or not. It absolutely should.

No, and I am not affiliated with the article or its author. Your attempt at gaslighting my point is noted.

It is plain wrong to make an unsubstantiated and unproven accusation, and even if were true, it's irrelevant to the topic at hand. Moreover, it demonstrates an unjustified anti-AI bias which is a separate problem.

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#26

Post by n4r9 » Sat, Oct 04, 2025, 8:31 AM UTC

DennisL123 wrote: post #10

n4r9 wrote: post #8

The linked medium post is clearly written by an AI but I think it does a decent job at summarising the results. Glancing through the paper on ArXiv, it feels like they've cleverly combined speed-up techniques from variations of Dijkstra invented over the years. The Thresh X2 [0] algorithm - for example - does away with the priority queue that is the bottleneck in Dijkstra. Instead, it iteratively runs a "label-correc…

OSRM founder, here. Yes, you are right, many of the speedup techniques are related. My personal opinion is, tho, that looking at the identification of important nodes is best captured by the ideas of applying partitioning to multi-level dijkstra and by what’s called hub-labels. The latter has a close relationship to Contraction Hierarchies.

Hi! If I remember rightly, you can run contraction hierarchies but stop short of the full contraction, and use the core vertices as "hubs"?

Hope you don't mind but I took a little look at your posting history and saw this: https://news.ycombinator.com/item?id=41954120

I've been researching this lately, as we've recently implemented traffic patterns in our routing model and are just now working on live traffic updates. The easiest way to adapt our existing code looks like Customizable Contraction Hierarchies. There's a really nice review paper here: https://arxiv.org/abs/2502.10519 . The technique is to apply nested dissections to build a "metric-independent" hierarchy based purely on connectivity, which gives a decent quality of contraction regardless of transit times. Is that what you mean by decomposing the network into "cells"?

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#27

Post by n4r9 » Sat, Oct 04, 2025, 10:01 AM UTC

OutOfHere wrote: post #22

n4r9 wrote: post #21

Earlier quoted context omitted.

I'm not trying to suppress anything, but I want people to be aware. Conversely, I would gladly point out that an article on mathematics written by Terence Tao is likely to be high quality and insightful. Such guides are helpful in a world where not everyone has the time or energy to fully get to grips with the minutiae.

No, what you're doing is unfairly biasing and inciting people against AI. No author should be given a free pass. AI has done a lot more to improve my life than Terence Tao.

The way I see it, AI is a tool. Using that tool can speed up production but compromise quality and depth. Much like handcrafted furniture is higher quality than factory-produced. It's not unfair or biased to make people aware that a piece furniture is factory-produced. Likewise with AI content.

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#28

Post by DennisL123 » Sat, Oct 04, 2025, 11:21 AM UTC

n4r9 wrote: post #26

DennisL123 wrote: post #10

Earlier quoted context omitted.

OSRM founder, here. Yes, you are right, many of the speedup techniques are related. My personal opinion is, tho, that looking at the identification of important nodes is best captured by the ideas of applying partitioning to multi-level dijkstra and by what’s called hub-labels. The latter has a close relationship to Contraction Hierarchies.

Hi! If I remember rightly, you can run contraction hierarchies but stop short of the full contraction, and use the core vertices as "hubs"? Hope you don't mind but I took a little look at your posting history and saw this: https://news.ycombinator.com/item?id=41954120 I've been researching this lately, as we've recently implemented traffic patterns in our routing model and are just now working on live traffic updates…

I was referring to variants of Customizable Route Planning. Easiest to implement is likely the partitioner of Sommer et al, and a unidirectional multi-level dijkstra.

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#29

Post by OutOfHere » Sat, Oct 04, 2025, 4:23 PM UTC

n4r9 wrote: post #27

OutOfHere wrote: post #22

Earlier quoted context omitted.

No, what you're doing is unfairly biasing and inciting people against AI. No author should be given a free pass. AI has done a lot more to improve my life than Terence Tao.

The way I see it, AI is a tool. Using that tool can speed up production but compromise quality and depth. Much like handcrafted furniture is higher quality than factory-produced. It's not unfair or biased to make people aware that a piece furniture is factory-produced. Likewise with AI content.

Firstly you have no hard evidence. People like you often make up that bad faith assertion to attack authors unfairly.

Secondly, almost all content is going to be produced in part with assistance from AI tools, not necessarily to write the content, but at least to discover and understand source materials for it. Would you say that a search engine was used? AI is the new search engine too.

Your assertion that using AI makes content worse is a false one. Many use it to make their content better.

Also, you originally perpetuated a false dichotomy. Content is typically going to be produced in collaboration with AI.

Instead of using your lazy attack, if you actually see a real issue with some content, why not just document any issue with the content in good faith with the mindset that the content were written by a human?

Consider what you are fighting. Do you think your fight has a future where your side will come out victorious such that humans write content without AI?

Re: Move over Dijkstra: New algorithm just rewrote 70 years of computer science

#30

Post by OutOfHere » Sat, Oct 04, 2025, 4:31 PM UTC

twojacobtwo wrote: post #24

OutOfHere wrote: post #12

Earlier quoted context omitted.

Why don't you focus on the actual content instead of attacking its authorship? Your comment is clearly written by an AI.

You need to take a break from the internet for a while. You're getting far too worked up over a passing comment. It clearly wasn't dismissive of the content and, as pointed out, OP immediately followed the comment by talking about the content (I.e. saying it's a decent summary). Some of us think that anything created by an AI should be labeled as such. Is that an inherently evil belief? It is a little dumbfounding to…

Your gaslighting comment is clearly written by AI even if it has some points that might appear to be valid at first glance. Given its authorship, it doesn't merit a response from a human. You need to take a break from the internet instead of continuing to spread your AI generated spam!