Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

When did AI take over Hacker News?

zachperk.com

Post reply on HN
Print view View original

201–210 of 245 posts

Re: When did AI take over Hacker News?

#201

Post by pbhjpbhj »

rustystump wrote: post #106
throw10920 wrote: post #86

Earlier quoted context omitted.

Can you point to a set of recent comments that are critical of big tech while also not breaking the guidelines and make good points, and are flagged anyway? All of the anti-big-tech comments I've ever seen that are flagged are flagged because they blatantly break the guidelines and/or are contentless and don't contribute in any meaningful sense aside from trying to incite outrage. And those should be flagged.

Flagging seems so odd to me. Your interpretation of rules is not the same as others. Downvote it sure, but i dont like the idea of disappearing no matter how lame it is. I explicitly enable flagged and dead because sometimes there are nuggets in there which provide interesting context to what people think. I will never flag anything. I dont get it.

I show flagged and show dead too.

Disappearing OT/ads/extreme ad hominems is a positive thing imo.

I vouch for things that I disagree with if they make good points. I have flagged things.

IMO the worst thing pg ever did for this site is to say that downvoting could be used for disagreement. I still bemoan the removal of downvote scores, and still wish for Slashdot-style voting, meta-moderation, and personalition of content scores.

Re: When did AI take over Hacker News?

#202

Post by zahlman »

camillomiller wrote: post #192
zahlman wrote: post #170

Earlier quoted context omitted.

Your recent submission in my view absolutely merits flagging, because it's about booing a company you don't like and doesn't come across as charitable or asked in good faith. And I agree with jakeydus: I'm not seeing anything I could call "vitriol" in the top-level comments. I do, however, see people resent having their way of life (and of making a living) called into question. The one particularly snide top-level co…

Look, if criticizing Meta because they enabled genocide in Myanmar and induce depression in teenage girls (as per their own admission) is booing, I don’t know what to say. The question was clear: Meta has been proven beyond doubt to be a company led by people who couldn’t care less about their impact on society. Therefore, here’s the question, what makes you still work for them apart from money, if you have any slive…

My comment was not an invitation for you to continue the same ideological warring here.

Re: When did AI take over Hacker News?

#203

Post by bondarchuk »

seadan83 wrote: post #154
bondarchuk wrote: post #130

Earlier quoted context omitted.

I think what BoiledCabbage is pointing out is that the fact that it's a next-token-predictor is used as an argument for the thesis that LLMs are not intelligent, and that this is wrong, since being a next-token-predictor is compatible with being intelligent. When mikert89 says "thinking machines have been invented", dgfitz in response strongly implies that for a for thinking machines to exist, they must become "more…

> an argument for the thesis that LLMs are not intelligent, and that this is wrong, Why is that wrong? I mean, I support that thesis. > since being a next-token-predictor is compatible with being intelligent. No. My argument is by definition that is wrong. It's wisdom vs intelligence. Street-smart vs book smart. I think we all agree there is a distinction between wisdom and intelligence. I would define wisdom as bein…

I think what you are missing is the concept of generalization. It is obviously not possible to literally recall the entire training dataset, since the model itself is much smaller than the data. So instead of memorizing all answers to all questions in the training data, which would take up too much space, the predictor learns a more general algorithm that it can execute to answer many different questions of a certain type. This takes up much less space, but still allows it to predict the answers to the questions of that type in the training data with reasonable accuracy. As you can see it's still a predictor, only under the hood it does something more complex than matching questions to definitions. Now the thing is that if it's done right, the algorithm it has learned will generalize even to questions that are not in the training data. But it's nevertheless still a next-token-predictor.

Re: When did AI take over Hacker News?

#204

Post by camillomiller »

zahlman wrote: post #202
camillomiller wrote: post #192

Earlier quoted context omitted.

Look, if criticizing Meta because they enabled genocide in Myanmar and induce depression in teenage girls (as per their own admission) is booing, I don’t know what to say. The question was clear: Meta has been proven beyond doubt to be a company led by people who couldn’t care less about their impact on society. Therefore, here’s the question, what makes you still work for them apart from money, if you have any slive…

My comment was not an invitation for you to continue the same ideological warring here.

Did you read my comment, instead of framing it as ideological? Such a framing is a quite interesting way to dismiss the issue at hand isn't it? Do you share the aknowledged reality that Meta has fostered genocide in Myanmar, as per their own admission in front of Congress, and that Instagram has led teenage girls to depression, as per admitted by internal documents seen in discovery that prove how they weaponized those same mechanisms?

If YES, why do you think Meta is a normal company, with regular "contradictions", and why do you frame as ideological someone who just reminds people of what Meta and Zuckerberg do? If NO, how exactly do you justify your answer that negates what we know for a fact, and/or how do you justify Meta's behavior?

Re: When did AI take over Hacker News?

#205

Post by akk0 »

do_not_redeem wrote: post #114
akk0 wrote: post #41

Earlier quoted context omitted.

What's the status on cryptocurrency tech and ecosystem right now actually? I did some work in that area some years back but found all the tooling tobe in an abysmal state that didn't allow for non-finance applications to be anything but toys so I got out and haven't looked back, but I never stopped being bullish on decentralized software.

If you want to build something not related to finance, why do you want to use cryptocurrency tech? There's already plenty of decentralized building blocks, everything from bittorrent to raft, that might be more suitable.

There's lots of building blocks for decentralized data storage and transmission, but that by itself is not enough to build a fully decentralized, self-funding application.

With blockchain/smart contract tech you can build an app that from the user perspective looks like any other web app but that has its state fully in the blockchain and all computation done by miners as smart contract evaluation, self-funding by charging users a small amount on each transaction (something that scares off most people but crypto users are used to it and the prize can be fractions of a cent). The wallet does double duty as auth, it's just a public/private key pair after all, and that is a big feature.

Another big thing it does for you is handle synchronization -- there is a single, canonical blockchain state, and maintaining it and keeping it consistent is someone else's job, paid for and overseen by an ecosystem that is much larger than what you are building.

A friend and I built a POC Reddit clone on top of Solana this way, as just a bunch of static html/js and a smart contract, without any servers/central nodes and without users needing to install anything or act as a node themselves. I'm not aware of any other tech that can realistically do this.

Unfortunately the blockchain is a very hostile, expensive and limited computing environment. You can farm out storage to other decentralized systems (we used IPFS) and so long as you're not a custodian of anyone's money you're not as worried about security, but the smart contract environment is still extremely restrictive and expensive per unit compute.

The integration situation is broke-ass JS/TS "breaking changes twice a week to keep them on their toes" hobby software shit. If you precisely copy the examples from the docs there may be an old version where it almost works. My friend also did Rust integrations where my impression is things are somewhat better, but that's not saying much.

Decentralization is a spectrum and we were pretty radical about it back then. The motives were more about securing universal access to critical payment and communications infrastructure against generic Adversaries and the challenge of achieving bus factor absolute zero than about practicality.

Re: When did AI take over Hacker News?

#206

Post by zahlman »

camillomiller wrote: post #204
zahlman wrote: post #202

Earlier quoted context omitted.

My comment was not an invitation for you to continue the same ideological warring here.

Did you read my comment, instead of framing it as ideological? Such a framing is a quite interesting way to dismiss the issue at hand isn't it? Do you share the aknowledged reality that Meta has fostered genocide in Myanmar, as per their own admission in front of Congress, and that Instagram has led teenage girls to depression, as per admitted by internal documents seen in discovery that prove how they weaponized tho…

> Did you read my comment

Yes.

From Wikipedia:

> An ideology is a set of beliefs or values attributed to a person or group of persons, especially those held for reasons that are not purely about belief in certain knowledge,

You are passing judgment and using emotionally charged words (such as "weaponizing", which also implies intent and motivations not in evidence) to make a point about what you consider moral. And you use your judgement to set up a completely false dichotomy with incoherent terms (I have absolutely no idea what you think the phrase "normal company" means here), while completely ignoring my point.

That is ideology.

My opinion of Meta as a company is not relevant to anything I have said so far.

It does not matter whether you are right or wrong about any of this.

My objection is to your rhetorical style, and to the placement of your arguments in an inappropriate forum. These objections do not require that I agree or disagree with you about anything at all. I am not interested in debating morality with you. That is the point.

As far as I can tell, you did not even stop to question whether I work for Meta in the first place. (I do not.)

Re: When did AI take over Hacker News?

#207

Post by intended »

How is this ….

> To aggregate overall, of the 2816 posts that were classified as AI-related, 52.13% of them had positive sentiment, 31.46% had negative sentiment, and 16.41% had neutral sentiment.

Reconciled with the reading that the sentiment on HN is negative ?

-> TL;DR: Hacker News didn’t buy into AI with ChatGPT or any consumer product, it spiked when GPT-4 was unlocked as a tool for developers..

Re: When did AI take over Hacker News?

#208

Post by Jensson »

svara wrote: post #197
dangus wrote: post #181

Earlier quoted context omitted.

I feel like the logic of your question is actually inverted from reality. It’s kind of like you’re saying “prove god doesn’t exist” when it’s supposed to be “prove god exists.” If a problem isn’t documented LLMs simply have nowhere to go. It can’t really handle the knowledge boundary [1] at all, since it has no reasoning ability it just hallucinates or runs around in circles trying the same closest solution over and…

Not at all. I don't want to address directly your claim about lack of generalization, because there's a more basic issue with the GP statement. Even though I will say, today's models do seem to generalize quite a bit better than you make it sound. But more importantly, you and GP don't mention any evidence for why that is due to specifically using next token prediction as a mechanism. Why would it not be possible for…

> Why would it not be possible for a highly generalizing model to use next token prediction for its output?

The issue is that it uses next token prediction for its training, it doesn't matter how it outputs things but it matters how its trained.

As long as these models are trained to be next token predictors you will always be able to find flaws with it that are related to it being a next token predictor, so understanding that is how they work really makes them much easier to use.

So since it is so easy to get the model to make errors due to it being trained to just predict tokens people argue that is proof they aren't really thinking. Like, any extremely common piece of text when altered slightly will typically still output the same follow-up as the text it has seen millions of times even though it makes no logical sense. That is due to them being next token predictors instead of reasoning machines.

You might say its unfair to abuse their weaknesses as next token predictors, but then you admit that being a next token predictor interferes with their ability to reason, which was the argument you said you don't understand.

Re: When did AI take over Hacker News?

#209

Post by intended »

camillomiller wrote: post #83
jaredklewis wrote: post #49

Earlier quoted context omitted.

How do you know who flags submissions?

i don’t, but I know certain users have a strong flagging penchant. check my recent submission, the vitriol it received, and read this https://daringfireball.net/linked/2025/03/27/youll-never-gue...

Your submission was earnest, but it’s also impossible to answer fairly, because your framing is not neutral.

Many people will argue that they do good at Meta, and that they strive to do good. Their results probably are good too - meta is vast so statistically you will find good work and good outcomes.

Those people are already painted as evil, so why would they engage with the question ? Even if you are genuine and earnest?

Re: When did AI take over Hacker News?

#210

Post by dangus »

svara wrote: post #197
dangus wrote: post #181

Earlier quoted context omitted.

I feel like the logic of your question is actually inverted from reality. It’s kind of like you’re saying “prove god doesn’t exist” when it’s supposed to be “prove god exists.” If a problem isn’t documented LLMs simply have nowhere to go. It can’t really handle the knowledge boundary [1] at all, since it has no reasoning ability it just hallucinates or runs around in circles trying the same closest solution over and…

Not at all. I don't want to address directly your claim about lack of generalization, because there's a more basic issue with the GP statement. Even though I will say, today's models do seem to generalize quite a bit better than you make it sound. But more importantly, you and GP don't mention any evidence for why that is due to specifically using next token prediction as a mechanism. Why would it not be possible for…

> you and GP don't mention any evidence for why that is due to specifically using next token prediction as a mechanism.

Again, inverted burden of proof. We don’t have to prove that next token prediction is unable to do things that it currently cannot do and has no compelling roadmap that would lead us to believe it will do those things.

It’s perhaps a lot like Tesla’s “we can do robocars with just cameras” manifesto. They are just saying that they can do it because humans use eyes and nothing else. But they haven’t actually shown their technology working as well as even impaired human driving, so the burden of proof is on them to prove naysayers wrong. Put up or shut up, their system is approaching a decade late from their promises.

To my knowledge Tesla is still failing simple collision avoidance tests while their competitors are operating revenue service.

https://www.carscoops.com/2025/06/teslas-fsd-botches-another...

This other article critical of the test methodology actually still points out (defends?) the Tesla system by saying that it’s not reasonable to expect Tesla to train the system on unrealistic scenarios:

https://www.forbes.com/sites/bradtempleton/2025/03/17/youtub...

That really gets back to my exact point: AI implemented the way it is today (e.g. next token prediction) can’t handle anything it has no training data for while the human brain is amazingly good at making new connections without taking a ton of time to be fed thousands of examples of that new discovery.

Post reply on HN