October 21, 2022

s13e12: Standards; The Bird in the Room

0.0 Context Setting

It's Friday, 21 October, 2022 in Portland Oregon and the air is finally good, green and healthy to breathe outside. Well. Healthier.

School -- for our kids, at least -- was cancelled yesterday, and I've realized that our kids are more likely to have school cancellations due to bad air than snow. (Or more pandemics, I suppose).

---

1.0 Some Things That Caught My Attention

Standards

If you subscribe to inevitable progress whether brute force on one end or more elegantly and strategically, intentionally designed and implemented on the other, then what I'm thinking about in this bit probably doesn't matter.

What I'm thinking about is the fact that for a bunch of people, machine learning-based image generation (aka DALL·E and successors) is good enough despite the fact that it cannot (like most humans, to be fair), do hands well.

On the one hand you can think that it's just a matter of time before nVidia or whomever is able to deliver sufficient teraflops for something to learn how to reproduce hands that fit whatever training criteria exist for "the correct type of hands implicitly implied by someone's instructions".

On the other hand, you could look at this and say: wait, where did we get to "you know what, fucked-up hands¹ are totally okay, I didn't need hands that looked like hands, this is good to go"? I'm not thinking about this in terms of being reactionary about ML image generation, more a sort of ivory tower pondering "what does it say about us that we're totally fine with the rough idea of a hand and not an accurate depiction of a hand".

I get that it depends mostly on context: if the overall intent of the image generation is creative exploration then hands, of course, are something you can fix in post. If, however, you are say an internet columnist who writes a regular newsletter and you're using ML image generation for the hero image of your story and it's about e-bikes or bikes, then you probably want an image that depicts a bike and not what most humans would draw as a bike from memory if they had not really looked at a bike and practiced drawing one recently.

So, I guess, if you're just sketching and roughing something out, then sure, maybe you don't need five digits on a hand or for the thumb to be opposable if that's not the point. It is a bit weird though, and perhaps that's just something we end up calibrating on. Or, you know, ML image generators get better at hands.

---

The Bird in the Room

I noticed yesterday that the general plan, or as much a plan can be discerned, is that Elon completes his purchase of Twitter a week from today, on Friday 28 October.

This realization came via a Washington Post story that dropped during the day, that Musk was planning on cutting 75% of Twitter's workers² and hidden amongst the story was the deal close date.

Next week! I thought we had at least 3 more months and apparently I'm not the only one.

I wrote about this only a few days ago in s13e07³, where I thought that all the overlapping communities on Twitter would be unbundled and split off into their own separate spaces. And that they'd indeed be in separate spaces and by that I meant separate applications/services/websites: they all have different feels. Slacks are not a Discord is not a mailing list is not a subreddit is not a Twitter Community is not a Facebook Group and so on.

So to the answer of "where will we go?" I think the answer is "lots of other places", and it depends on who your we is, and what your we is at any time.

I counted at least 11 shared interest groups that I belong to / show up in that might be distinct communities amongst who I follow on Twitter, and by "communities" I mean areas of interest where I have regular mutual follower relationships and actually interact every now and then.

It's easy to say that social media doesn't work and point to the giants in the room and the difficulties they face. That's the wrong issue. Social media -- that is, online community, is absolutely fine. It works, as I've always said, in subreddits that are tended to, in Facebook Groups that are actively run by people who have clear community values, it works in standalone communities that have coalesced over time (see, forever and ever: Metafilter).

The failure mode of a social network, where failure is "turns out to be too much of a toxic, abusive environment full more of noise than signal" is more a function of size, I think, than it is of the type of platform. Of course, certain platforms are predicated towards certain sizes.

One thing I think is this: mass social media doesn't work right now. One of the reasons why I think it doesn't work is because it needs to exist in an environment of unknown-lateness-capitalism, where it's subject to other systemic incentives and disincentives. This is subtly different from the argument that "you can't say social media doesn't work, all you can say is that we haven't tried", but I pull on the "haven't tried" thread, and the largest incentives towards "not trying [properly]" are "because it has to exist in the economic world we have right now".

Some notes:

• Small communities are more likely to be sustainable and healthy than larger communities
• Communities live/die/exhibit healthy behavior as a direct result of their governance, which results in norms, and so on. Mumsnet, in the UK, is a large online community, bigger than Metafilter, a good-sized subreddit, I'd think, and is by all accounts in some areas a shitshow, but it hasn't died
• The various chans are still around*, barring legal issues and social shunning and shaming, regardless of size, mainly because of their governance, even if you think their reason for existing is shit. They have not (well, they have a bit) collapsed.

Where it doesn't quite work is the attempt to produce a modern agora or public square for hundreds of millions of people with different backgrounds, points of view, moral and ethical systems, combined with incentives to influence the behavior of the people in that square by those willing or able to do so. A naive reading of this might be that we're not ready as a collective species for this type of public square, and that attempting to squish everyone into a public square when we're not ready or "mature" enough, which includes the economic environment and its dis/incentives once it hits a certain size or scale. Polities already want to restrict online speech and behavior to bring them more into line with their historical, polity-specific norms according to wherever they are on this supposed arc of progress that is continually being bent (which, I would note, requires someone to actually give a shit and do the bending).

Another thought I have builds on the suspicion/verging sometimes on contempt of Metcalfe's law, that the value of a network increases proportional to the square of the number of connected users to the system. I criticised how our professional community in general treats this law back in 2018⁴ (it all goes back to how you define "value", right?).

So let me try to bring this back:

• Mass social media public squares above a critical size of, say, hundreds of millions of useres, right now cost too much in terms of implementing quality of experience, given the economic environment.
• Mass social media public squares are valuable in that they are the current best opportunity for naive serendipity and "meeting new people", see anecdata about "Twitter has been great for meeting people in whatever community" and, bizarrely, the chance of being exposed to adjacent points of view in, I don't know, your vector space.
• When I say "current best opportunity" this doesn't mean they're good at it, this instead means that they're the least-bad-way of meeting other people on a mass scale. This "how bad they are at it" is constantly in flux because at a high level it's a result of how The Algorithm [sic] is tweaked in terms of what you're exposed to, and product decisions such as how shares/retweets/likes and so on and other signals are exposed in your feed.
• "Moderation" is unscaleable and doesn't work at "acceptable cost" (to the extent that it's even been tried) at polity-scale. I think you might need to go back to the AOL/Compuserve/Prestel/Minitel era for polity-scale (albeit tiny, tiny uptake) social-ish networks that at first order would be inclined to reflect the governance and high-level moral/ethical values of most of their members.
• The economic (and political) incentives to create and manage polity-scale (ie more than, say, 250 million aggregate regular users or, thinking out loud, more a function of proportion of users in a particular nation state) are super high! (As an aside, one reason why I think Murdoch et al haven't acted like they have pressure to "have a social thing" lately is that their existing mass media networks are super good at providing content that someone else is going to distribute. They have their own problems in terms of ad revenue and so on, but if you're Murdoch and you're as much in it to influence, you're doing fine)

So what works? If mass social media just handwaves doesn't or can't work above a certain threshold of nation-state penetration, what are we supposed to do, or what might it look like for us?

• Small networks, very loosely joined
• Small networks that cater to different kinds of places and affordances
• A loss of no-effort serendipity and "meeting new people"; because we're not "all in the same place" anymore. There's less "hey, everyone's in this bar, let me introduce you to my friend over here". If we want to meet new people, it's going to require more active human curation [sic], i.e. personal introductions or connection-making than we've been used to for the last few years. This seems fine and a bit like some sort of parental "if it's important then you're going to have to work for it". This isn't to say there's no room for algorithmic You Might Be Friends With, but let's just say that that recommendation system is more on the crapshoot/harmful end if we want to be deliberate about the quality of introduction.
• Firebreaks. I wrote about firebreak protocols back in 2018⁵, the idea that actually it might be a good idea to have regulators in place to prevent the rapid spread of content. Or, that die-off be rapid. I actually don't think we really have a problem with rapid die-off, we seem to be exhibiting a sort of spiking attention where things burn bright and die out before terminally online people seek the next hit of synchronous mass common experience and hot-take and joke generation. Maybe some communities die and that's fine and you get regrowth that's kind-of-but-not-exactly the same. Maybe you do blow away your network and recreate it every now and then. Hell, what is it like to move to a new town. We've essentially seen this behavior when someone is socially shunned out of the public square and then they stomp off and either build their own village which lives or dies based on at the moment their sheer fucking incompetence, or, you know, just... limps along, but is elsewhere now. They get to recreate a network, and it's kind-of-the-same-but-not.
• Small networks created explicitly by humans with vouching and referral. Or, if you want to be shitty about it and needlessly reductive "safe spaces". These small networks right now are most likely little Discords or Slacks with private invitations and the implicit or explicit instruction that what happens there stays there. They're shared interest, and they may or may not be sustainable, but the differentiator is that they're not designed to grow indefinitely.
• Paying for it. Metafilter costs about five bucks. Communities on Slack and Discord exist, as pointed out by a friend, on the sufferance of either VC or "because someone else has a giant CRM that a lot of people use". They should cost money. Here's some models: that bar you meet up with your friends at or that pub or whatever, they provide a relatively well curated space that's clean enough and even better complies with regulations regarding providing a safe space that, you know, won't burn the fuck down, or if it does start burning, you can actually escape and in exchange you buy things like drinks from them. In the absence of buying drinks from them, we're (sigh) "paying with our data" or acting as loss-leaders. The point is, you have been paying about five bucks a month (or more) at some point in your life to have a place to hang out with your friends, and at this point in our technological development, running a server somewhere and essentially (sorry, Slack) a good-enough IRC server for however-many-of-your-friends-times-five-bucks-a-month is easily enough. So. We could pay for it. And yeah, I know, it's just another bill and there's so many of those right now and many of them are critical especially right now, and that's shitty. So hey, do what you can. Fuck, I don't know. Some government systems even provide community spaces that groups of people can hire and use! I mean, imagine that! Imagine getting a non-extractive place for a community to meet online because it's good for society in general? It's almost like you'd be advocating for community-owned network access!
• Not everyone has to be an organizer, there just have to be enough organizers. It's a lot of work to manage a community. But it's fortunate, I think, that out of all the temperaments we exhibit as people, one of those is "likes to connect people and gets energy and derives enjoyment when doing a good job". People will want to do this, not everyone has to do this.
• The in-between is us. There's a slight joke, I think, amongst my friends who are members of many, many Slacks or so on, which is meme arbitrage: you see something funny in one Slack, invariably in a channel for sharing good tweets or eyerolling stuff or whatever, and it is not in another Slack yet, so you post it there. You may not have been the person who "discovered" it, but you're essentially passing something along. It is weird, for me at least, that this feels kind of dirty? And yet it shouldn't, I don't think? In mass social media public squares, "passing things along" became quote-tweeting, first manual (see! a baseline human behavior!), and then built into the product, and then used as as signal to (ostensibly) keep us engaged in the mainline feed, but then in an opaque way with "trending topics", or "things that your friends liked" whether or not you gave a shit. If your friend really liked it and wanted you to know, they would've, you know, told you. So. We would go back, more, I think, to telling people things and it would be fine for some people to be connectors and messengers. None of this is new! There's probably a bad Malcolm Gladwell article about it, we know this stuff!
• Clear governance, rules, and being willing to call people on shit and throw them out. Your pub, if it is a good one, will throw people out. Your bartender, if they are good, will notice if someone is hassling you and do something about it and keep you safe. You want to go somewhere that's like that, and there are people who run communities like that because they care about their friends and there's increasing awareness of what to look for and willingness to enforce it. Because this is much more clearly your space, and you have agency in how it's run, rather than the public square, where all you have right now is the option to not listen, to a certain degree, to assholes.

Anyway. Twitter is supposedly encountering a phase shift in the way a whale encounters the ground of a supposedly desolate planet next Friday and this was, in some part, what has caught my attention about it.

---

That's it. It's Friday. The adaptation of William Gibson's The Peripheral is out today on Amazon Prime Video which on some degree is laughs weakly what can you do hilarious because of course The Jackpot. Jackpot Enables Popular Media About Jackpot. Excellent, well done humanity.

How are you?

Best,

Dan