I'm guessing he means there's no algorithms that decide what you see and what you don't, you just see everything from everyone you follow. But I don't use Tumblr so I'm not 100% sure.
They still do. There are algorithms to fetch whatever will be shown on your feed. They may be as straightforward as fetching a user's follow list and displaying their posts chronologically in reverse. That's still an algorithm.
I'm sure Gaiman means specifically the kind of "smart" feed algorithm like FB etc employ. That said, I really dislike this trend of equating "algorithms" to something wicked. I write them for a living, and I'd rather we kept writing and employing them.
> They may be as straightforward as fetching a user's follow list and displaying their posts chronologically in reverse. That's still an algorithm.
Correct, it's absolutely an algorithm. Crucially, it involves extra steps to make it scalable without needing a materialized inbox. Or at least that's how Tumblr's reverse-chron dashboard worked from 2009ish through at least 2018, maybe still today.
There were a bunch of optimizations in there to keep the queries fast, and minimize the number of rows that need to be examined and sorted. One key step involved cross-referencing the blogs you follow against their latest post timestamp. For example, say you're fetching posts 11-20 on the reverse-chron feed. The worst-case is that posts 1-20 all come from different blogs, so you can sort the list of followed blogs by latest-post-timestamp, and then only examine/sort posts from the top 20. (That's a slight simplification; you actually need to look at the timestamp from post 10, and examine posts from however many followed blogs posted since then, plus 10 more. I spent months of my life tuning this stuff back in 2011-2012...)
That all said, in the context of dashboard feeds, "algorithm" often just means the opposite of reverse-chron. But even then, the "no algorithms" statement is completely incorrect! Tumblr has a "Best Stuff First" setting which controls whether or not your feed is reverse-chron. For a while in late 2017 (iirc) onwards, this setting was even enabled by default... maybe still is?
"Algorithm" is used in everyday language to mean "system of determining which content you get shown and the order it gets shown in." The "absence" of an "algorithm" in this context just means "you are shown the content generated by people you follow in the order that it is published."
I have a hard time believing you didn't understand the original post. Obviously the above does not mean there are no algorithms in a literal sense.
I'm amused that Tumblr is becoming a frontrunner as Twitter replacement.
I'm not 100% convinced by it, but it _seems_ like a better proper replacement than Mastodon (sorry FOSS people, but... centralized logins is a big deal compared to federated spaces).
At least, its the closest thing I've seen discussed that seems to do a similar job to Twitter. Sharing comments / images / short form video with a large community of communities with a somewhat trusted login / identity
Mastodon doesn't fix that, because Alice@instance1.com doesn't necessarily mean you're talking to Alice. It could be a malicious Alice. Maybe the "real" Alice is Alice@instanceAlpha.com. Twitter used to have blue-checkmarks for this, but that identify management seems to be getting destroyed by the current chaos, and people are rightfully wondering how to get a trusted identity for their discussions.
> Mastodon doesn't fix that, because Alice@instance1.com doesn't necessarily mean you're talking to Alice. It could be a malicious Alice. Maybe the "real" Alice is Alice@instanceAlpha.com.
Okay, I'll bite; how do you know that the Tumblr user actualice12345 is the real Alice?
I'm not a Tumblr or Twitter user. But it seems like Tumblr has a degree of identity verification going on with its users.
-----
Even Reddit has these kinds of simple checks to make "Ask me Anything" a more legitimate-feeling experience. When a famous reporter posts a picture of Reddit-account name + face for people to see and do on an AMA, it helps verify the identity.
(And if that's too much for the AMA user, having a trusted moderator do that kind of check can serve as a good substitute)
The way it's always worked; she can publicly post the link to her Tumblr on her Alice.com homepage, or you confirm with her in private outside of Tumblr.
I'm not familiar with Mastodon, but this is the generic mechanism for linking identity since the earliest web days; as long as names are hard to spoof it's independent of any specific service.
For fiction and chatting, no. For eyewitness accounts or providing evidence, yes, provenance matters.
Also, a famous name (or at least, one well-known to you) has a lot of embedded trust associated with it that was earned elsewhere, which is why impersonation matters.
People don't do large-scale public chats with celebrities over email.
Think Reddit AMAs, or media interviews and the like. When someone says "This is reporter X talking with Alice (picture of Alice + voice + etc. etc.)", it establishes identity that people trust. I guess people could fake this these days, but building up these networks of trust for worthwhile interaction has grown more and more important recently.
Twitter's blue checkmark kinda-sorta served this purpose, at least before it died last week.
-----
If there was a widespread "Ask me anything" email discussion organized over emails, this identity verification problem would also need to be solved. Since Twitter is an "always ongoing" discussion with various celebrities, the blue checkmark was kinda-sorta the verification you needed.
Of course it does, email from stuff lime suppport@google.com is often used for phishing. Not to mention, email is mostly 1:1, so it has completely different discovery/identity validation properties than a social network that's normally n:n.
Even for things like newsletters, you subscribe to a newsletter. It's true that mailing lists exist, but they don't scale and require curation precisely to establish the identity of each participant.
Email has this problem all the time! [1] Real estate companies have to give customers cybersecurity advice [2] because phishers pretend to be the agent and send incorrect addresses for wire transfers.
We should not call it verification IMHO. The malicious user can create random website and add the rel=me. The "green" link makes impression that the user is "verified".
(I would prefer to have some chain of trust like pgp but with a better UX.)
I would hope the design doesn't lead to that effect, because it's not like Twitter where there's a blue tick mark that means "verified" v.s. "non verified".
Instead, there is an area on your profile page with links, and the links themselves that are verified are shown in with a tick - which should indicate "this user is verified as being linked to this website", not "this user is a verified user".
I'd be interested in research that shows that the Mastodon approach confuses people. My hunch is that it doesn't, but it's just a hunch.
I don't care if I talk with Alice or with Bob, I'm only interested in what "Alice" or "Bob" have to say, i.e. what they actually write down/share, I'm not at all interested in their identities.
Because of that the coming apart of the whole blue-checkmark thing is really great, as it changes us all back to "Alice" and "Bob" and, hopefully, in so doing it will put the impetus back on what we are actually saying and not on our Alice-ness or Bob-ness.
Fully anonymous just leads to people building tons of fake accounts. You could be talking to 20x people, but they're all Bob who is just obsessed with proving his point and making it look like his point is popular.
A degree of identify verification, to mitigate bots, spammers, mass posters, and other such issues is at minimum for discussion on today's internet.
> You could be talking to 20x people, but they're all Bob who is just obsessed with proving
If Bob goes full Pessoa and creates 20 different heteronomies I'm all for it, but unfortunately that never quite happens and you get the gist of Bob1, Bob2... Bob20 being all the same person pretty fast.
> and other such issues is at minimum for discussion on today's internet.
This place seems to function decently well, at least for now, and I didn't even have to give them my email address.
Because its fully anonymous chat/discussion. I'm wondering if you have experience with anonymous chat where identities don't matter.
Unlike you, I have experienced it back in the day. So I think I have a thing to say or two about anonymous chat, the tactics that occur on those forums and how meme culture spreads, how the community uses memes to "defend" themselves and establish a group identity, etc. etc.
If you don't care about identities, you'll likely build a system that turns into 4chan (or something like it). That's all I'm saying.
When you say things like:
> I'm not at all interested in their identities.
Well, that reminds me of 4chan. That's all. If you really didn't care, there's a forum out there that works like that. I do suggest you try it out for a bit and figure out why its culture is the way it is.
> This place seems to function decently well
Because moderators like @dang come around here and ban accounts that engage in these activities.
Mastodon allows for the servers to "ban" other individuals or even servers that are misbehaving, so its not fully anonymous or unmoderated. Its not as trusted as a central server though (like Twitter was). Hard to see where everything will end up.
I have not been active on chat groups for 20 years now, give or take, but I have been an anonymous user of both this website and of reddit for 15 years now, with varying levels of success. Were these 2 websites to ask me for my real identity I would just ditch them (even though, if one looks close enough, one can get to my real name by following some links to my projects I posted some time ago)
> If you really didn't care, there's a forum out there that works like that
I'm not really interested in the identity of people posting here, the same goes for reddit. As I said, I have been quite happy with this state of things for 15 years now. They might decide to change it going forward, but, then again, there's always other anonymous forums somewhere on the web.
Related to this, interesting how things have changed compared to, when was it?, 2012-2014?, i.e. when Google was pushing for the "real identity" stuff for their Google+ users. I remember the backlash from back then, almost none of it has remained.
You're arguing this on a discussion forum that works just fine without identity verification (since it relies on long-term reputation for pseudonymous accounts).
This discussion site has moderators actively patrolling for throwaway accounts, and uses shadow-banning to discourage them from messing up our discussions.
We need those kinds of assurances to have decent discussions on any site today. I'm not 100% sure if Tumblr has it or how Twitter worked either. But identity is important.
Maybe not the _SPECIFIC_ identity of who you or I am. But enough of an identity to say that you (or I) am not double-making accounts, astroturfing opinions or the like.
That's what I mean by long-term pseudonymous accounts. My point is that persistent identity (which we are both talking about here) is not the same as verified identity. As long as there's a way to build reputation on the site, readers can distinguish newbie accounts and impostors from long-running community members. You don't need a mechanism to check anyone's ID.
>I'm amused that Tumblr is becoming a frontrunner as Twitter replacement.
Is replacement needed for Twitter? Will Elon really destroy it? If that's the case Twitter's replacement needs to go to the Twitter roots of status updates and not repeat the same mistakes.
Fixing the authenticity problem for institutions is as simple as running a mastodon server on your own domain, which would cost less per user than a blue check.
What a bunch of manipulative bastards. Instead of giving people the freedom to curate their timelines, and educating them how to do so, they prefer to give them spoonfed feeds because "users are lazy". Everyone is lazy, it's basic energy conservation principle, but if people stopped motivating each other to learn we d still be living in caves. Calling people "Normal people" just shows how entitled these devs are
I saw a bunch of posts on the tumblr frontpage without being logged in. How are these determined, if not by some kind of algorithm? I made a new account and I saw a bunch of posts on my feed, apparently personalized based on my past surfing, without having followed anyone. Again, same question.
I have nothing against "algorithms", but why is the CEO of the company that owns tumblr reblogging someone making claims about tumblr that are just not true?
I've finally been able to follow someone (I tried this already yesterday, but their posts didn't show up in my feed) and I see their posts now, but interspersed with "check out these blogs". It's a million times more obnoxious then twitter, and on twitter you have "latest tweets" mode where actually see only the tweets from people you follow, in reverse chronological order and nothing else.
Surprisingly little known, but Twitter also has a strict-timeline mode. It's great! Tap the pair of stars in the upper-right to bring up a menu to turn it on.
The supposed post-Twitter situation is entirely imaginary and typical of Twitter culture: hysterical, performative, not really rooted in any fact whatsoever.
But for the sake of argument, let's go along with the idea that a major shift is happening. The thing is, everything people found great about Twitter, they will not find in alternatives like Mastodon or Tumblr.
Those networks are organic, more personal and bubble-like, and relatively peaceful. Which is healthy, but entirely the opposite of Twitter which is algorithmic amplification of hot takes, outrage, division and the many engagement farming tactics that come along with it.
On an organic reach platform, that behavior doesn't work. Which is a steep change. To a Twitter user, such a platform will feel broken. It's quiet, peaceful, and they're not ranking up followers or influence with low effort hot takes.
Not only that, the hot takes may get them banned. Twitter users have been using toxicity as a tool for so many years that they've normalized behavior and language that is widely unacceptable outside Twitter.
Twitter users escaping Twitter do not want something completely different. They want Twitter but without Elon Musk. Which is largely performative. But you can't replicate the Twitter experience outside of Twitter.
Note how none of the big accounts are leaving. A true departure means deleting your account. Nobody does it. They keep rambling on about Mastodon yet fail to get even 5% of their followers there. They keep coming back to say how great Mastodon is, but never actually leave Twitter. Some may deactivate their account but never delete it. They know very well that a big following on Twitter cannot be recreated anywhere else.
The other thing the alternatives cannot replace is Twitter's news factor. Whether it is sports, coverage of war, whatever is happening in the world, it's happening in real time on Twitter. Absolutely never on Mastodon.
48 comments
[ 4.8 ms ] story [ 125 ms ] threadI'm sure Gaiman means specifically the kind of "smart" feed algorithm like FB etc employ. That said, I really dislike this trend of equating "algorithms" to something wicked. I write them for a living, and I'd rather we kept writing and employing them.
Correct, it's absolutely an algorithm. Crucially, it involves extra steps to make it scalable without needing a materialized inbox. Or at least that's how Tumblr's reverse-chron dashboard worked from 2009ish through at least 2018, maybe still today.
There were a bunch of optimizations in there to keep the queries fast, and minimize the number of rows that need to be examined and sorted. One key step involved cross-referencing the blogs you follow against their latest post timestamp. For example, say you're fetching posts 11-20 on the reverse-chron feed. The worst-case is that posts 1-20 all come from different blogs, so you can sort the list of followed blogs by latest-post-timestamp, and then only examine/sort posts from the top 20. (That's a slight simplification; you actually need to look at the timestamp from post 10, and examine posts from however many followed blogs posted since then, plus 10 more. I spent months of my life tuning this stuff back in 2011-2012...)
That all said, in the context of dashboard feeds, "algorithm" often just means the opposite of reverse-chron. But even then, the "no algorithms" statement is completely incorrect! Tumblr has a "Best Stuff First" setting which controls whether or not your feed is reverse-chron. For a while in late 2017 (iirc) onwards, this setting was even enabled by default... maybe still is?
I have a hard time believing you didn't understand the original post. Obviously the above does not mean there are no algorithms in a literal sense.
In any case, it doesnt really seem like he knows anything about this sort of stuff.
I love Neil Gaiman's works, but he's no technology wizard - he's an author trying to surf this insanity we call social media.
And that's OK. There are algorithms, yet his advice is still sound.
I suppose we can blame a combination of ignorance and marketing.
I'm not 100% convinced by it, but it _seems_ like a better proper replacement than Mastodon (sorry FOSS people, but... centralized logins is a big deal compared to federated spaces).
At least, its the closest thing I've seen discussed that seems to do a similar job to Twitter. Sharing comments / images / short form video with a large community of communities with a somewhat trusted login / identity
Mastodon doesn't fix that, because Alice@instance1.com doesn't necessarily mean you're talking to Alice. It could be a malicious Alice. Maybe the "real" Alice is Alice@instanceAlpha.com. Twitter used to have blue-checkmarks for this, but that identify management seems to be getting destroyed by the current chaos, and people are rightfully wondering how to get a trusted identity for their discussions.
Okay, I'll bite; how do you know that the Tumblr user actualice12345 is the real Alice?
https://pbs.twimg.com/media/DXC71NFX4AAUPX3?format=jpg&name=...
I'm not a Tumblr or Twitter user. But it seems like Tumblr has a degree of identity verification going on with its users.
-----
Even Reddit has these kinds of simple checks to make "Ask me Anything" a more legitimate-feeling experience. When a famous reporter posts a picture of Reddit-account name + face for people to see and do on an AMA, it helps verify the identity.
(And if that's too much for the AMA user, having a trusted moderator do that kind of check can serve as a good substitute)
Also, a famous name (or at least, one well-known to you) has a lot of embedded trust associated with it that was earned elsewhere, which is why impersonation matters.
Interestingly, email seems to not have that problem at all.
Think Reddit AMAs, or media interviews and the like. When someone says "This is reporter X talking with Alice (picture of Alice + voice + etc. etc.)", it establishes identity that people trust. I guess people could fake this these days, but building up these networks of trust for worthwhile interaction has grown more and more important recently.
Twitter's blue checkmark kinda-sorta served this purpose, at least before it died last week.
-----
If there was a widespread "Ask me anything" email discussion organized over emails, this identity verification problem would also need to be solved. Since Twitter is an "always ongoing" discussion with various celebrities, the blue checkmark was kinda-sorta the verification you needed.
Even for things like newsletters, you subscribe to a newsletter. It's true that mailing lists exist, but they don't scale and require curation precisely to establish the identity of each participant.
[1]: https://www.proofpoint.com/us/threat-reference/email-spoofin...
[2]: https://www.rate.com/resources/how-to-avoid-phishing-scams-a...
It lets you add verified links to your profile - which could be another rel=me supporting site (such as GitHub) or your own personal domain.
I wrote some notes about that here: https://til.simonwillison.net/mastodon/verifying-github-on-m...
(I would prefer to have some chain of trust like pgp but with a better UX.)
Instead, there is an area on your profile page with links, and the links themselves that are verified are shown in with a tick - which should indicate "this user is verified as being linked to this website", not "this user is a verified user".
I'd be interested in research that shows that the Mastodon approach confuses people. My hunch is that it doesn't, but it's just a hunch.
I don't care if I talk with Alice or with Bob, I'm only interested in what "Alice" or "Bob" have to say, i.e. what they actually write down/share, I'm not at all interested in their identities.
Because of that the coming apart of the whole blue-checkmark thing is really great, as it changes us all back to "Alice" and "Bob" and, hopefully, in so doing it will put the impetus back on what we are actually saying and not on our Alice-ness or Bob-ness.
Fully anonymous just leads to people building tons of fake accounts. You could be talking to 20x people, but they're all Bob who is just obsessed with proving his point and making it look like his point is popular.
A degree of identify verification, to mitigate bots, spammers, mass posters, and other such issues is at minimum for discussion on today's internet.
Why would I do that?
> You could be talking to 20x people, but they're all Bob who is just obsessed with proving
If Bob goes full Pessoa and creates 20 different heteronomies I'm all for it, but unfortunately that never quite happens and you get the gist of Bob1, Bob2... Bob20 being all the same person pretty fast.
> and other such issues is at minimum for discussion on today's internet.
This place seems to function decently well, at least for now, and I didn't even have to give them my email address.
Because its fully anonymous chat/discussion. I'm wondering if you have experience with anonymous chat where identities don't matter.
Unlike you, I have experienced it back in the day. So I think I have a thing to say or two about anonymous chat, the tactics that occur on those forums and how meme culture spreads, how the community uses memes to "defend" themselves and establish a group identity, etc. etc.
If you don't care about identities, you'll likely build a system that turns into 4chan (or something like it). That's all I'm saying.
When you say things like:
> I'm not at all interested in their identities.
Well, that reminds me of 4chan. That's all. If you really didn't care, there's a forum out there that works like that. I do suggest you try it out for a bit and figure out why its culture is the way it is.
> This place seems to function decently well
Because moderators like @dang come around here and ban accounts that engage in these activities.
Mastodon allows for the servers to "ban" other individuals or even servers that are misbehaving, so its not fully anonymous or unmoderated. Its not as trusted as a central server though (like Twitter was). Hard to see where everything will end up.
> If you really didn't care, there's a forum out there that works like that
I'm not really interested in the identity of people posting here, the same goes for reddit. As I said, I have been quite happy with this state of things for 15 years now. They might decide to change it going forward, but, then again, there's always other anonymous forums somewhere on the web.
Related to this, interesting how things have changed compared to, when was it?, 2012-2014?, i.e. when Google was pushing for the "real identity" stuff for their Google+ users. I remember the backlash from back then, almost none of it has remained.
We need those kinds of assurances to have decent discussions on any site today. I'm not 100% sure if Tumblr has it or how Twitter worked either. But identity is important.
Maybe not the _SPECIFIC_ identity of who you or I am. But enough of an identity to say that you (or I) am not double-making accounts, astroturfing opinions or the like.
Is replacement needed for Twitter? Will Elon really destroy it? If that's the case Twitter's replacement needs to go to the Twitter roots of status updates and not repeat the same mistakes.
> normal people
What a bunch of manipulative bastards. Instead of giving people the freedom to curate their timelines, and educating them how to do so, they prefer to give them spoonfed feeds because "users are lazy". Everyone is lazy, it's basic energy conservation principle, but if people stopped motivating each other to learn we d still be living in caves. Calling people "Normal people" just shows how entitled these devs are
I have nothing against "algorithms", but why is the CEO of the company that owns tumblr reblogging someone making claims about tumblr that are just not true?
[1] https://www.tumblr.com/explore/trending
But for the sake of argument, let's go along with the idea that a major shift is happening. The thing is, everything people found great about Twitter, they will not find in alternatives like Mastodon or Tumblr.
Those networks are organic, more personal and bubble-like, and relatively peaceful. Which is healthy, but entirely the opposite of Twitter which is algorithmic amplification of hot takes, outrage, division and the many engagement farming tactics that come along with it.
On an organic reach platform, that behavior doesn't work. Which is a steep change. To a Twitter user, such a platform will feel broken. It's quiet, peaceful, and they're not ranking up followers or influence with low effort hot takes.
Not only that, the hot takes may get them banned. Twitter users have been using toxicity as a tool for so many years that they've normalized behavior and language that is widely unacceptable outside Twitter.
Twitter users escaping Twitter do not want something completely different. They want Twitter but without Elon Musk. Which is largely performative. But you can't replicate the Twitter experience outside of Twitter.
Note how none of the big accounts are leaving. A true departure means deleting your account. Nobody does it. They keep rambling on about Mastodon yet fail to get even 5% of their followers there. They keep coming back to say how great Mastodon is, but never actually leave Twitter. Some may deactivate their account but never delete it. They know very well that a big following on Twitter cannot be recreated anywhere else.
The other thing the alternatives cannot replace is Twitter's news factor. Whether it is sports, coverage of war, whatever is happening in the world, it's happening in real time on Twitter. Absolutely never on Mastodon.