zlacker

[parent] [thread] 22 comments
1. lars_f+(OP)[view] [source] 2024-02-02 16:40:06
> While I have no reason to doubt Daniel's good faith, it's hard to believe that HN users would be tired of LLM-related news.

I am, so you can believe it. But: I don't flag things that I'm tired of.

replies(8): >>add-su+J3 >>nemoth+N3 >>isopro+97 >>IshKeb+D8 >>dang+b9 >>dpkirc+5e >>brucet+4i >>cayman+0z
2. add-su+J3[view] [source] 2024-02-02 16:53:53
>>lars_f+(OP)
If I could, I'd hibernate until such time as I didn't have to hear about generative AI anymore.
replies(2): >>SoftTa+N5 >>kjkjad+s6
3. nemoth+N3[view] [source] 2024-02-02 16:54:01
>>lars_f+(OP)
There was some heavy handed moderation decision that moot made, can't remember what, but he enforced it by saying "One man's shitpost is another man's board culture". I think about that a lot when it comes to moderation because people tend to assume everyone in the community is just like them; and really only moderators have a gauge on how saturated certain can be.

It's also why I don't like the "free speech at all costs" meme that gets thrown around when $corporation bans $person_i_like. Every community needs moderation and it's often a thankless job that feels like nothing is being done at all when it's being done right.

replies(1): >>captai+ef
◧◩
4. SoftTa+N5[view] [source] [discussion] 2024-02-02 17:02:02
>>add-su+J3
I agree, I don't find it very interesting.
◧◩
5. kjkjad+s6[view] [source] [discussion] 2024-02-02 17:05:20
>>add-su+J3
It took about 10 years for the crypto headline hysteria to taper but it might have been only because ai is now the big annoying thing to shoehorn into everything. Monkeys pawl will curl and you will emerge out of your hibernation to more disgust at whatever the next annoying thing will be.
replies(2): >>IshKeb+U8 >>isopro+W8
6. isopro+97[view] [source] 2024-02-02 17:09:02
>>lars_f+(OP)
I make my money building things with LLMs and even I am tired of reading about them
7. IshKeb+D8[view] [source] 2024-02-02 17:16:09
>>lars_f+(OP)
Yeah me too and I also wouldn't flag them. I flag things that are false or misleading or just especially stupid.
◧◩◪
8. IshKeb+U8[view] [source] [discussion] 2024-02-02 17:17:19
>>kjkjad+s6
I think the difference is that AI is definitely useful and here to stay.

Crypto was mostly scams or pie in the sky ideas that will never work. It will stick around for money laundering & buying drugs but that's about it.

replies(1): >>smaude+3c
◧◩◪
9. isopro+W8[view] [source] [discussion] 2024-02-02 17:17:27
>>kjkjad+s6
Also, it's very clear from the messaging and breathless hype that the NFT grifters packed uo their stuff and moved over en masse to the GenAI space
replies(1): >>projec+5w
10. dang+b9[view] [source] 2024-02-02 17:18:39
>>lars_f+(OP)
> it's hard to believe that HN users would be tired of LLM-related news

For sure many are. This happens with every Major Ongoing Topic (MOT) and LLMs are way beyond a MOT [1]. The hivemind tires of repetition extremely quickly [2]. The trick is to try to separate wheat from chaff, where 'wheat' means the stories that bring Significant New Information (SNI) [3] and 'chaff' means the follow-up and copycat stories, which are legion [4].

It's important to understand are that there's a wide spectrum of opinion about this stuff. If you imagine a slider with "allow zero posts about $TOPIC" at one end, and "allow all posts about $TOPIC" at the other end, pretty much every user would slide it to a different position. This is true for every $TOPIC and especially for the biggest ones.

Frontpage space is the scarcest resource HN has [5] and every reader has a different 'signature' of preferences that they would like to see (or not see) there. This means not only that it's impossible to satisfy everybody, but that it's impossible to fully satisfy anybody—because nobody's 'signature' is perfectly matched on the front page, and (lest any of you be thinking of this quick riposte) certainly not the mods'!

[1] https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

[2] https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...

[3] https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...

[4] https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

[5] https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...

replies(1): >>maxbon+Ec
◧◩◪◨
11. smaude+3c[view] [source] [discussion] 2024-02-02 17:32:55
>>IshKeb+U8
Eh? I'm not pretending digital coins or dubious "tokens" were all particularly useful, but crypto (in the sense of cryptography) has been around for decades and is definitely here to stay...

The people who made coins and tokens bad for society are doing the same thing with GenAI...

Both are useful and both come with huge problems. Neither one is some panacea or a sustainable get-rich-quick scheme (obviously, both people in "crypto" and in "GenAI" are getting rich, but neither are going to lead to some sort of great societal good).

replies(1): >>sfink+8i
◧◩
12. maxbon+Ec[view] [source] [discussion] 2024-02-02 17:35:24
>>dang+b9
Have you ever considered writing a book about what you've learned about moderation and community?

You seem to have developed these concepts pretty extensively. Seeing you break down this terminology whets my appetite to hear from you in long form.

replies(1): >>dang+gf
13. dpkirc+5e[view] [source] 2024-02-02 17:40:58
>>lars_f+(OP)
I will occasionally flag things that will result in discussions that are always the same because I'm tired of them. Stories about tipping at restaurants or Trump or Biden, for example -- literally every argument for or against has been made and there's nothing new or interesting to say. But I'm more likely to hide them.
replies(2): >>Pareto+ph >>em500+Ll
◧◩
14. captai+ef[view] [source] [discussion] 2024-02-02 17:46:11
>>nemoth+N3
Why does seeing moot quoted suddenly make me feel old.
◧◩◪
15. dang+gf[view] [source] [discussion] 2024-02-02 17:46:33
>>maxbon+Ec
I'd be pretty miserable doing that, but one of these years I'd like to condense the past explanations into something a bit more definitive and put them up as sort of glosses on the site guidelines. I imagine most of those HN Search links I'm constantly posting could be replaced by a link to some sort of canonical paragraph on the topic.
◧◩
16. Pareto+ph[view] [source] [discussion] 2024-02-02 17:55:53
>>dpkirc+5e
Why would one ever flag stories they believe will result in the same useless discussions rather than just hiding them?

I think I've only ever flagged one or two instances of spam personally.

17. brucet+4i[view] [source] 2024-02-02 17:58:06
>>lars_f+(OP)
LLMs are like crypto, where scams and scam-adjacents are everywhere.

I am the biggest local ML advocate you will find. My 3090 is either running Yi 34B queries or other experiments all day, my job is with local LLMS... But I am totally OK with heavy handed AI-related moderation. I dont want the sea of AI grifters to have a single second on the HN front page.

◧◩◪◨⬒
18. sfink+8i[view] [source] [discussion] 2024-02-02 17:58:18
>>smaude+3c
> but crypto (in the sense of cryptography) has been around for decades and is definitely here to stay...

But that's not the sense under discussion. "Crypto"=cryptography lost the language war and was completely supplanted by "crypto"=cryptocurrency. I really wish the word could regain its original and useful meaning, but it's too late now.

Ironically, "I work in crypto" went from meaning something useful to society to meaning being a parasite on society, and you'd best not accidentally use the phrase expecting people to understand it to mean the original thing (cryptography).

(Yes, not all uses of cryptocurrency are a parasitic detriment. But if you happen to be working on actually useful stuff and we meet socially, then please be very quick about saying that you work at doing something with cryptocurrency or blockchain that is intended to provide actual benefit. If you just say "I work in crypto", I will excuse myself at the first opportunity.)

replies(1): >>smaude+Nn
◧◩
19. em500+Ll[view] [source] [discussion] 2024-02-02 18:14:22
>>dpkirc+5e
I would also include the periodic Monty Hall re-post (everything that ever comes up in the discussions can be found in the Monty Hall problem wikipedia page).

And also pretty much any article about inflation.

◧◩◪◨⬒⬓
20. smaude+Nn[view] [source] [discussion] 2024-02-02 18:23:14
>>sfink+8i
> "Crypto"=cryptography lost the language war and was completely supplanted by "crypto"=cryptocurrency.

On the timescale of the past 4-5 years, you are correct about the popular usage.

However, if cryptocurrency continues to recede from the public eye, then in another 4-5 years I think "crypto" will no longer mean "cryptocurrency".

Understanding both the current lexicon and the "archaic" and "recently archaic" uses of the term I hold is both useful and pertinent to being able to communicate effectively. Which is why I immediately clarified, I'm talking about the 40+ year definition of the term, not the current whimsical linguistic fad.

◧◩◪◨
21. projec+5w[view] [source] [discussion] 2024-02-02 19:01:39
>>isopro+W8
Was at a networking event recently and "I was in the crypto/NFT space but I'm now pivoting to GenAI" was by far the most common way people introduced themselves.

At least it made it easy to figure out who I didn't need to talk to.

22. cayman+0z[view] [source] 2024-02-02 19:17:27
>>lars_f+(OP)
I'm sick of LLM-related news. I'm fascinated by the technology and the progress, but for every one article about something novel, there are dozens rehashing the same points about social impact, bias, deepfakes, plagiarism, etc. These topics are of some interest to me, but the vast majority of the articles bring nothing new to the table and are reactionary responses to the latest infraction.
replies(1): >>dang+EQ1
◧◩
23. dang+EQ1[view] [source] [discussion] 2024-02-03 05:46:18
>>cayman+0z
I believe that is close to what the median HN reader feels: interested by the significant new developments, fatigued by the endless incremental updates, and grossed out by the hypemeisters.
[go to top]