Thread
Day 2 of complexity science for @elonmusk. Today's topic:
Stability, Resistance, and Robustness: will twitter collapse?
Buckle up, we're going on another 🧵
Stability, Resistance, and Robustness: will twitter collapse?
Buckle up, we're going on another 🧵
Twitter is inherently a complex system. Complex systems are defined by interactions between individual units such that the functioning or behavior of the whole is greater than the sum of its parts. A watch is complicated, a power grid is complex.
Complex systems are very hard to reason about through argument alone. Conway's game of life is a classic example: simple rules define the birth and death of pixels.
You're not going to talk your way into knowing if adding/deleting pixels or adjusting rules is going to make it last longer or shorter, change the behavior, or exhibit a desired pattern. You've gotta run simulations and try to discern general principles.
Coming back to Twitter, it is a complex system in multiple ways. The software that it depends on involves interlinked processes that depend on one another in complex ways. The behavior of users, the actions of advertisers.... you name it: It's complex.
Many complex systems exhibit stability: They persist in some state for a period of time, ideally a functional one. Twitter has been stably serving up hot-tweets and surveillance capitalism for some years now.
Perturbing a complex system can jeopardize its stability, testing its robustness. If it's sufficiently robust, it'll absorb the blow without changing its state. A robust power grid can handle the tea kettles turning on all at once, across England
en.wikipedia.org/wiki/TV_pickup
en.wikipedia.org/wiki/TV_pickup
However, robustness is finite and if you push it too far you then have to hope the system is resilient---returns to its stable state after being perturbed. When the power goes, supply chains fail, or the market crashes, how long til things go back to normal?
The trick to managing a complex system that is currently stable is to avoid massive perturbations, especially if it is overconnected. Here's an intro, complete with pre-financial-collapse worry about overconnectedness from derivatives.
www.nature.com/articles/451893a
www.nature.com/articles/451893a
Elon has already established that he wants to make this network even more connected. The code too, focusing on retaining the developers who wrote the *most* lines of code. He's pushing them to role out features, all of which will undermine stability.
On top of this, he's trying to cut infrastructure costs. That means turning things off that other things might depend on. Shut off one power station, no big deal. Shut off half and you don't get half the power.. you get none. Cascading failure.
www.reuters.com/technology/musk-orders-twitter-cut-infrastructure-costs-by-1-bln-sources-2022-11-03/
www.reuters.com/technology/musk-orders-twitter-cut-infrastructure-costs-by-1-bln-sources-2022-11-03/
Not content with jeopardizing the software infrastructure, the human element is being cut as well. These are the people who know how to fix things when they're broken.... they are the resilience.
www.nbcnews.com/business/business-news/twitter-layoffs-elon-musk-fires-staff-how-many-rcna54799
www.nbcnews.com/business/business-news/twitter-layoffs-elon-musk-fires-staff-how-many-rcna54799
This is not just the code... it's the human element as well. A big uptick in hate speech, misinformation, or harassment can drive users and advertisers away. Great way to do this would be by laying off trust and Safety, human rights, etc..
What you *really* wouldn't want to do is cause your highly connected nodes to leave. I'd do this by forcing them to pay money, or removing status. Something like forcing them to pay for verification.
On top of this, his ad sales are dropping due---in no small part---to his behavior and decision-making. This drop in ad revenue might be resilient, but it's also possible they just shift their business elsewhere and don't come back.
So will twitter fail? If I wanted to keep an eye peeled for failure I'd look for a phenomenon known as "critical slowing down". Systems about to transition will take longer and longer to get back to normal. Ad sales, latency, etc..
www.pnas.org/doi/10.1073/pnas.1406326111
www.pnas.org/doi/10.1073/pnas.1406326111
Will it? This happened to Digg after a large shift. Digg went from influencing the flow of internet traffic (the "digg effect") to a ghost town after a big redesign caused cascading failures across its code, then userbase and ultimately advertisers.
en.wikipedia.org/wiki/Digg
en.wikipedia.org/wiki/Digg
Myspace, Friendster, Google's many bids, ICQ, AIM... the graveyard of social networks is full of companies that perturbed things a bit much and fell apart at the seams.
I can't tell you if Twitter will fail, but I can say he's doing everything I could think of break it or any complex system.
If it does, let's hope it doesn't take down any of the other complex systems it's attached to (e.g., journalism, science, politics).
If it does, let's hope it doesn't take down any of the other complex systems it's attached to (e.g., journalism, science, politics).
Let's end with conway's game of life. If you had a game that had been running for years, would you delete half the dots all at once in hope of making it better?
Typo apologies, as is usual. Errands to run!
Mentions
See All
Carl T. Bergstrom @CT_Bergstrom
·
Nov 4, 2022
Last week—no, that was yesterday!!—@jbakcoleman posted an excellent thread about Musk's collective, cybernetic super-intelligence fantasy and why Twitter doesn't work like that. Today he's back with a second must-read thread about complexity and the Twitter ecosystem.