A very interesting article. It seems like the Community Notes system is built on (in my opinion) a very mathematically sound model. It seems far less susceptible to biased "mob actions" than a conventional upvote/downvote system, even if it's not impervious to such actions. "Bias" exists in the eye of the beholder, so it's unreasonable to expect any content rating system to be perfectly unbiased.
Community Notes and crypto share a common thread of "complicated math that makes it both extremely powerful and totally opaque to 95% of people." Honestly, that's a pretty weak relationship for how much page-space it got in the writeup. Looks like Vitalik is writing a crypto blog, so I guess you gotta appeal to your readers. Great read nonetheless.
The association with crypto wasn't merely about the math aspect, but also the principles of open source verification of the algorithms and credible neutrality.
True, but that "95% of people" part is key. The auditability of open source doesn't matter (matters less?) to non-technical audiences because they don't have the knowledge, time, and/or wherewithal to audit the code themselves. As a result, they have to defer to a third party to do the auditing.
At best, a truly independent third party auditor has a mild philosophical incentive to recommend an open-source product over a closed-source one of similar quality. In practice, it's far more common for third party "stamps of approval" to be meaningless gold stars purchased by corporations to aid in marketing their product. A great example of this is the iF Design Award.
I really like community notes. It's been a while in Brazil they are being used to check politicians who always lie in their post and try to rewrite history.
Its like a small note of shame in their posts, and it seens so far they been using it in politicians in all sides of the political spectrum.
> deep learning vs crypto is a clear divide of rotators vs wordcels. the former offends theorycel aesthetic sensibilities but empirically works to produce absurd miracles. the latter is an insane series of nerd traps and sky high abstraction ladders yet mostly scams
Can anyone translate this from meme to English please?
Also, it would seem to be a very anti-crypto statement for Vitalik to be posting?
No, it's common knowledge most publications on crypto in general are scams, and Vitalik has always been critical about money grabbing schemes. Vitalik's project Ethereum has been hurt a lot by "shit coins" (alternative cryptocurrencies that have money grabbing as their sole objective) as if there had been a magical way for regular users to become aware of which crypto technologies add something material to the space and which not, the space as a whole would be a lot healthier.
The other part of that sentence about crypto is a funny self jab though, apparently he's conscious about the research rabbit holes and mathematical abstractions the Ethereum team went through to get for example staking to work.
Had Ethereum not been near the top of every crypto exchange, and Vitalik not been somewhat of a known person, how difficult would it have been for the average crypto-interested person to tell Ethereum's "insane nerd traps and sky high abstraction ladders" from the scams?
Shape rotator = someone who can rotate shapes in their head, a proxy for IQ. Someone with good mathematical or STEM skills and who feels unvalued for it.
Wordcel = someone good with words/rhetoric who feels like their skill in this field isn't valued enough / it doesn't get them far enough.
> Can anyone translate this from meme to English please?
"ChatGPT, invent some jargon to make people who've bought in feel like insiders, while making ungullible people^W^W I mean haters keep a safe distance."
Community Notes will likely have some of the same pros and cons of wikipedia.
Community Notes are not written or curated by some centrally selected set of experts; rather, they can be written and voted on by anyone ... It's not perfect, but it's surprisingly close to satisfying the ideal of credible neutrality...
Well, that's the optimistic pro. But the con is that if a particular demographic is more drawn to contributing to those notes (or comes to overwhelm Twitter itself), we will see the same problems of bias we see on Wikipedia in (say) social-cultural subjects which - whatever it results in - is certainly not "credible neutrality".
Wikipedia is WAY worse because they enshrine particular editors over particular sets of pages (esp. political ones) to make sure the edits only go the way those people want.
Anyone can review community notes and mark them as useful/not useful.
I don't know if community notes takes this into account, but you could weight bias according to how a contributor has voted in the past and medianize the output to prevent "dogpiling" or "brigading" from one side or the other.
>For a note to achieve a high intercept term (which is the note’s helpfulness score), it must be rated helpful by raters with a diversity of viewpoints (factor embeddings)
Which sounds like a pretty terrible idea in itself. At least to me, it would sound pretty chaotic if the edit that wins out on Wikipedia is the one that receives the most votes by other Wikipedia users, and if you wanted to get something on Wikipedia fixed, you'd need to gather enough people to support your fix over the previous version.
I've ran into issues with Wikipedia mods ignoring sources and going off whatever priors they had, but at least then I was able to just passive-aggressively berate them on the talk page to get them to bend.
The few that I have seen have been pretty good. I also like that they can be added to Ads, and many Ads running on Twitter right now have community notes because they're clearly saying one thing while the product does something else.
Yeah I’ve been surprisingly impressed. I remember the first thing I saw like this was Instagram’s fact checking which almost immediately turned into nonsense with people leaving fact check posts for obvious shitposts and memes.
The community notes on the other hand, seem to be well place, pretty informative, and in a few places surprisingly detailed.
Under no circumstances do you "gotta hand it to them", but Musk took the private beta, US-only Birdwatch, renamed it to something less cute (but ultimately a more pragmatic name), and launched it more broadly.
Someone alluded to this in the comments here, but this does strongly remind me of the Slashdot moderation method but much more sophisticated. That was my favorite site moderation method. Giving meta judgement/commentary on things felt like a privilege and not a given.
Readit News
Community Notes and crypto share a common thread of "complicated math that makes it both extremely powerful and totally opaque to 95% of people." Honestly, that's a pretty weak relationship for how much page-space it got in the writeup. Looks like Vitalik is writing a crypto blog, so I guess you gotta appeal to your readers. Great read nonetheless.
At best, a truly independent third party auditor has a mild philosophical incentive to recommend an open-source product over a closed-source one of similar quality. In practice, it's far more common for third party "stamps of approval" to be meaningless gold stars purchased by corporations to aid in marketing their product. A great example of this is the iF Design Award.
Its like a small note of shame in their posts, and it seens so far they been using it in politicians in all sides of the political spectrum.
It's always the same couple of lazy excuses.
Can anyone translate this from meme to English please?
Also, it would seem to be a very anti-crypto statement for Vitalik to be posting?
The other part of that sentence about crypto is a funny self jab though, apparently he's conscious about the research rabbit holes and mathematical abstractions the Ethereum team went through to get for example staking to work.
Had Ethereum not been near the top of every crypto exchange, and Vitalik not been somewhat of a known person, how difficult would it have been for the average crypto-interested person to tell Ethereum's "insane nerd traps and sky high abstraction ladders" from the scams?
"deep learning vs. crypto is a clear divide of math people vs creative people."
> the former offends theorycel aesthetic sensibilities but empirically works to produce absurd miracles
deep learning doesn't seem like it should work to people who are entrenched in theory, but somehow it produces great results.
> the latter is an insane series of nerd traps and sky high abstraction ladders yet mostly scams
crypto is full of interesting technical challenges but mostly produces scams.
https://knowyourmeme.com/memes/wordcel-shape-rotator-mathcel
Shape rotator = someone who can rotate shapes in their head, a proxy for IQ. Someone with good mathematical or STEM skills and who feels unvalued for it.
Wordcel = someone good with words/rhetoric who feels like their skill in this field isn't valued enough / it doesn't get them far enough.
"ChatGPT, invent some jargon to make people who've bought in feel like insiders, while making ungullible people^W^W I mean haters keep a safe distance."
Community Notes are not written or curated by some centrally selected set of experts; rather, they can be written and voted on by anyone ... It's not perfect, but it's surprisingly close to satisfying the ideal of credible neutrality...
Well, that's the optimistic pro. But the con is that if a particular demographic is more drawn to contributing to those notes (or comes to overwhelm Twitter itself), we will see the same problems of bias we see on Wikipedia in (say) social-cultural subjects which - whatever it results in - is certainly not "credible neutrality".
Anyone can review community notes and mark them as useful/not useful.
>For a note to achieve a high intercept term (which is the note’s helpfulness score), it must be rated helpful by raters with a diversity of viewpoints (factor embeddings)
https://communitynotes.twitter.com/guide/en/under-the-hood/r...
That's most of the article. If you're asking yourself that, you should read it
I've ran into issues with Wikipedia mods ignoring sources and going off whatever priors they had, but at least then I was able to just passive-aggressively berate them on the talk page to get them to bend.
The community notes on the other hand, seem to be well place, pretty informative, and in a few places surprisingly detailed.
It's nice to see some effort behind distributed moderation, but this is still too centralized to realize that dream.