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

A competitor crippled a $23.5M bootcamp by becoming a Reddit moderator

larslofgren.com

Post reply on HN
Print view View original

991–996 of 996 posts

Re: A competitor crippled a $23.5M bootcamp by becoming a Reddit moderator

#991

Post by DiggyJohnson »

seethedeaduu wrote: post #990
johnnyanmac wrote: post #840

Earlier quoted context omitted.

to be precise: >Communities and users that incite violence or that promote hate based on identity or vulnerability will be banned. is against TOC. You can talk about trans issues and offer reservations. You cannot say "trans people are a mental illness" or "trans are not people". That is clearly promoting hate and has nothing of substance to discuss. For a more explicit and current example, you can say "I don't think…

> You cannot say "trans people are a mental illness" or "trans are not people". That is clearly promoting hate and has nothing of substance to discuss. You definitelly can. There are plenty of big subreddits with posts like that, whose mods agree with.

Examples?

Re: A competitor crippled a $23.5M bootcamp by becoming a Reddit moderator

#992

Post by DiggyJohnson »

jibal wrote: post #733
DiggyJohnson wrote: post #720

Earlier quoted context omitted.

I can testify this is credible. You can prove this by trying it out yourself.

Your testimony isn't credible. You have made plainly false statements here.

What on Earth?

Re: A competitor crippled a $23.5M bootcamp by becoming a Reddit moderator

#994

Post by xpe »

oeirugjoiwejo wrote: post #592
binsquare wrote: post #161

This is oddly a case to signify there is value in an AI moderation tools - to avoid bias inherent to human actors.

> to avoid bias inherent to human actors. Do you understand how AI tools are trained?

Below I'll list each quote and rewrite them with elaboration to unpack some unstated assumptions. (These are my interpretations; they may be different than what the authors intended.)

> 1: This is oddly a case to signify there is value in an AI moderation tools - to avoid bias inherent to human actors.

"To the extent (1) AI moderation tools don't have conflicting interests (such as an ownership stake in a business); (2) their decisions are guided by some publicly stated moderation guidelines; (3) they make decisions openly with chain-of-thought, then such decisions may be more reliable and trustworthy than decisions made by a small group of moderators (who often have hidden agendas)."

> 2: Do you understand how AI tools are trained?

"In the pretraining phase, LLMs learn to mimic the patterns in the training text. These patterns run very deep. To a large extent, fine-tuning (e.g. with RLHF) shapes the behavior of the LLM. Still, some research shows the baseline capabilities learned during pretraining still exist after fine-tuning, which means various human biases remain."

Does this sound right to the authors? From what I understand, when unpacked in this way, both argument structures are valid. (This doesn't mean the assumptions hold, though.)

Re: A competitor crippled a $23.5M bootcamp by becoming a Reddit moderator

#995

Post by xpe »

oeirugjoiwejo wrote: post #592
binsquare wrote: post #161

This is oddly a case to signify there is value in an AI moderation tools - to avoid bias inherent to human actors.

> to avoid bias inherent to human actors. Do you understand how AI tools are trained?

Here is my high level take: most AI researchers I trust recognize that AI alignment is at least fiendishly hard and probably impossible. This breaks down into at least two parts. First, codifying the values of a group of people is hard and impossible to do neutrally, since many sets of reasonable desiderata fail various impossibility theorems, not to mention the practical organizing difficulties. Second, ensuring the AI generalizes and behaves correctly according to a based on supervised learning over a set of examples is likely impossible, due to the well-known problems of out-of-distribution behavior.

Of course, we can't let the perfect be the enemy of the better. We must strive to align our systems better over time. Some ways include: (a) hybrid systems that use provably-correct subsystems; (b) better visibility, vetting, and accountability around training data; (c) smart regulation that requires meaningful disclosure (such as system cards); (d) external testing, including red-teaming; (e) reasoning out loud in English (not in neuralese!); and more.

Re: A competitor crippled a $23.5M bootcamp by becoming a Reddit moderator

#996

Post by Montymio »

michaelnovati wrote: post #419
truetraveller wrote: post #377

Earlier quoted context omitted.

Thanks for your side. I hate fake reviews by competitors. But I read the article myself, and it seems exaggerated. It did read like a hit piece, and did feel ironic. This was before I even read your response. I don't know who's saying the truth, but it's never too late to better one's self. So that's the advice to myself and you.

I comment a lot about them and I have gotten annoyed every now and then when my tone was not professional, but most of critical comments about Codesmith center on The fact that their website has a giant banner saying from zero to mid-level engineer and I think that that's misleading and setting people up for failure in the software industry regardless of their outcomes or their talent. I'm open to hearing all sides o…

haven’t they literally launched thousands of graduates into mid level roles? Codesmith’s data backs it up, and i think that just pisses off Novati
Post reply on HN