Yesterday, I saw a tweet from Google AI guru François Chollet that I immediately knew was true, and it made my heart sink:
Here’s why I hated this tweet:
Before anything can be tackled with the software tools of ML and “big data,” it must be quantified.
Everything that is quantified is quickly gamified.
Everything that is gamified sucks, and sucks mainly in the same predictable ways.
Ergo, Chollet’s prediction reads to me as something like the following: in the future, everything will be a game that sucks and feels depressingly familiar.
Welcome to the Everything Game, where if you want to read, write, speak, buy, sell, date, marry, do science, play sports, or do anything else humans do, you will have no choice but to play. This game is soul-killing in its lameness and general uniformity, and suffocating in its total inescapablilty. But when software eats the world, we’ll all play this game for everything we do.
Here’s my first stab at outlining some basic characteristics of the Everything Game:
Everyone knows that the game is structured by a set of software-defined rules that do classification and ranking by taking in and evaluating various types of signals.
Everyone also knows that there are scores and metrics (of varying degrees of visibility to different players) that they must optimize in order to get certain desirable results from the game.
The knowledge of the deep structure of signals + rules + metrics, and how those three things combine to give rise to incentives, is distributed among the participants according to a power-law distribution most of the time. A few people know nearly all of the structure (especially the critical parts), and most people know some pieces of it.
Everyone knows about the power-law distribution and assumes (correctly) that the small club at the top of the distribution is using their superior knowledge to game the game itself at the expense of everyone else.
It’s difficult to tell “authentic” from “inauthentic” with respect to any type of signal, and in fact, this distinction is often more philosophical than practical or actionable. There’s always “playing the game” vs “gaming the game,” but those two modes of participation don’t always map to concepts of authenticity or inauthenticity, and indeed these concepts may not even have much utility.
The aforementioned power-law knowledge distribution is the equilibrium state, but will tend to break down as the more critical parts of the structure are uncovered by more people and winning strategies are crowded into.
When equilibrium breaks down because too many people are crowded into the same meta-game strategies, whoever has root has to step in and tweak the structure. This change to the structure creates a new temporary equilibrium, and onward the cycle goes.
In cases where the participants in the game are divided into are well-defined classes based on the nature of their participation (i.e., sellers vs buyers, or publishers vs readers), the class with less investment and less information (i.e. the buyers or readers) still has to know a certain amount about the way the other class is playing the game (and, ideally, the meta-game) in order to get good results.
Here’s what it always feels like to me to play the Everything Game:
I’m always either aware that I’m manipulating signals in order to make small, continuous optimizations for some target score, or I’m aware that others are doing it (and that it’s usually not in my interest for them to do so).
I’m always wondering what’s real or authentic, or what isn’t. Or to what degree something is authentic, or what “authentic” even means in this context.
I’m constantly trying to learn a bit more about the structure of the game, either because I need to play it better or because I need to be more aware of how others are playing it so that I can evaluate what I’m seeing.
I find myself constantly fighting against the desire to just quit all of the above efforts and settle into something really simplistic that requires less cognitive overhead and/or lets me check out of the game. E.g., if I’m shopping and worrying about fakes, I just buy directly from the manufacturer; if I see some claim of harm or trauma on social media, I assume the person is lying or exaggerating and keep scrolling.
There are a lot of human systems that look like the Everything Game, but what makes the Everything Game unique is that it’s mainly implemented in networked software and it happens at scale. Without all three of those things — the network, the software, and the scale — it’s not the Everything Game.
I’m sure I’ll want to revise the above as I think about this more, but in general, I think most or all of the above describes a pattern I find over and over again in online life, and that I expect to become more prevalent as everything moves into software.
The thing about all this that depresses me, is that this game is dumb and I’m tired of it. I’ve seen it too many times, and the thought of eventually being able to find no refuge from it anywhere is too much.
The problem of authenticity in social media
I want to zero in on one aspect of the Everything Game — the problem of authenticity — because it comes up in so many different ways in so many different domains.
Take social media, for instance. I’m a fairly sophisticated user of Twitter, so I’m aware of some of that particular game’s deep structure and of various strategies and exploits for optimizing things like engagement, likes, and follower count. So everything I do on the platform is informed to one degree or another by that knowledge.
Does the fact that I’m aware of the game and am never not playing it make my participation “inauthentic”?
Let’s take a concrete example. Earlier this week, I took a shot at Charlie Warzel, formerly of the NYT. Charlie and I have been friendly — he once interviewed me for his column — but I really disliked his characterization of the Basecamp situation. So I let him have it.
Charlie Warzel @cwarzelwho decides what is “political talk” at a company is...well...kind of important! https://t.co/IEjx6DnWzb
I did this out of genuine irritation, but also, I confess, because his account is a lot larger than mine, and I was hoping he’d respond to me (he did!). I’m aware of how the platform uses this kind of back-and-forth interaction to boost my own profile by surfacing the exchange in front of Charlie’s much larger audience. If his account were, say, half the size of mine, I’d have made a different calculation.
What makes this an example of the Everything Game, and not just some random example of social climbing, is the software. I know how the software works, and what signals it likes and dislikes, and I consciously calibrated my behavior in order to present the platform with a signal I wanted it to see.
Here’s another example — a set of examples, actually — of the Everything Game:
In this tweet, Kamath’s initial impulse is to throw some social software at the problem of distributing knowledge of and coordinating action around the errors in scientific papers. But he quickly realizes that he’d just be turning arXiv into a venue for the Everything Game.
A few of his followers suggest upvotes/downvotes, which of course doesn’t solve the problem at all — it just adds more signals to be manipulated. And then there’s this tweet:
Did this person post this because he’s super concerned about online abuse of women, or was he just trying to signal socially that he’s so concerned, or was he trying to signal socially that he’s so concerned because he is indeed so concerned and he thinks his signaling will help? Or was he trying to rack up some likes? Or was he trying to rack up likes in order to boost his profile so he can further promote the cause of stopping online harassment?
I think the answer is that none of these questions are important, because this kind of “authenticity” concern isn’t really compatible with the Everything Game.
Outside of a few narrow cases — mostly paid trolls, bots, or farms — it rarely makes sense to divide signals into “authentic” and “inauthentic” categories, because such divisions rely on certainty about internal mental states and intentions that even the signal-producers may not really have.
This is why accusations of “bad faith,” “virtue signaling,” “disinformation,” and “fake news” are signature moves in the Everything Game — because the game itself actively frustrates our human intuitions about authenticity by nakedly and explicitly reducing everything to a set of software-mediated experiences that consist of signals, rules, scores, and incentives. So you can always deploy these moves in the confidence that they either cannot be definitively countered or that the counter will be ignored as just another round of signal manipulation.
And when it comes to signals specifically, there are no “authentic” or “inauthentic” signals — there are only signals and noise. A signal may be miscategorized as negative or positive, or it may actually be noise, but “authenticity” isn’t really a category in information theory.
The authenticity problem in commerce
Now I’m going to think out loud about the authenticity problem in online commerce. A lot of this is half-formed, so I won’t be offended if you skip it.
First, a bit of backstory: I spent the previous ~2.5 years as a kind of professional shopper, by virtue of my time as Deputy Editor of a large prepper website. I would spend weeks putting together lists of products to buy, and then buying them and testing them for our reviews and guides.
Through this process I got a ton of intimate exposure to just how big a problem ecommerce fakery is. Fake reviews on ecommerce sites and in forums, fake ratings, paid reviews, counterfeit products, fake ads for real products... everything is fake everywhere you look. Here are two pieces I did specifically on the problem of fakes:
But sometimes, there are instances of inauthenticity that are like the social media examples, above — more philosophical than practically meaningful. Here’s a grab-bag of e-commerce examples:
I once saw a fake Amazon listing for a real solar panel, by which I mean I contacted the solar panel maker and found that this particular product was not sold in the US market, therefore the listing (which was an Amazon’s Choice listing for this item) was not posted by them or authorized by then; but likely the solar panel I’d get if I bought it was theirs.
There’s a popular brand of backpacks that’s often counterfeited, and I’m told that some of the counterfeits are actually after-hours, unauthorized production runs from the same factory in Asia that makes the official bags. The factory just does a bootleg shift for a maker who sells the bags online under another brand.
I once did a pretty deep dive on a company that I’m certain is a Chinese manufacturer that makes and sells survival/bushcraft fire-starting products to the US market under the pretense of being European. In this case, I guess you could say the brand is inauthentic, but the product is real?
When it comes to fakes of some downmarket, Big Box retailer hunting knives, it’s likely the case that the fakes are as good as the genuine article. Both are made of the same materials sourced from the same places, and are manufactured in the same region of China — maybe even in the same factory.
I read a recent viral talk about how the entire EU telecom industry is nothing more than a set of elaborate branding and financing operations wrapped around outsourced manufacturing, service, and customer support. From the piece:
So what have we outsourced? How bad is it? I made a sort of overview and it’s even worse than I thought. I know the telecommunication industry very well of course. And if you look within that world, they have outsourced the whole thing, and that includes the deployment of equipment. So the there’s actually no role for the telecommunication company there. They maybe will provide space for it. But also sometimes not that even, and the day to day maintenance, the system administration that has all outsourced, development, capacity building, administration, billing has all been outsourced.
So for example, invoicing, you’d think that sending out bills was core to telecommunications, but that also has been outsourced. So a typical telecommunications company, does not send equipment to its customers, does not design equipment, does not install equipment, does not maintain equipment does not send bills to its customers.
And you can wonder what do they do, and actually what they do is financing. But we’ll get to that later.
When companies structured this way claim to be in the telecommunications business via their marketing materials and corporate communications, are they being authentic or inauthentic? Does that question even matter? (Probably not.)
I suspect that when everything is mediated through networked software at scale, then everything becomes about signal manipulation and attempts to discover structure for the purpose of arbitraging your way into the top of the power law distribution.
This strikes me as no way to have a civilization, and I suspect that at some point we’re all going to find that out.