Profile picture
oliver beige @oliverbeige
, 29 tweets, 9 min read Read on Twitter
I think what he's trying to say is that life is a Boltzmann machine.…
And I tweeted this before I got to this point. @VladZamfir will hate this little trick...
FN: The "invention" was the objective function of a Boltzmann machine (invented by Hinton & Sejnovski), which turned out to be a potential function (invented by Monderer & Shapley). I didn't really invent it, I only applied it to P2P consensus convergence.
Modern life is just a massive multiplayer offline role-playing battle-of-the-sexes cum tragedy-of-the-commons game, choosing between what's best for all and what's keeping the peace.
The Industrial Revolution was about cooperation, the structured organization of supply and the emergence of mass production. The Digital Revolution is about coordination, the structured organization of demand and the emergence of individualized demand as drivers of value creation
The first problem of cooperation in production is longitudinal. Corporations as production functions solved the underlying prisoners dilemma. The problem of scale was solved by hierarchy, which worked as long as seniority reflected competence.
Information organizations don't work like that anymore. Competence sits anywhere in the hierarchy, and the problem becomes one of combining and matching skill sets to tasks. A problem of coordination.
I've discussed coordination on the demand side before. The problems of capturing coordination in a succinct model are, in order:
1. Scale
1. Structure and heterogeneity
1. Convergence
4. Equilibrium stability
5. Dynamic interaction & Bayesian updating
Economic theory has been very poor and slow at capturing any of them, largely bc of wrong model building. @glenweyl was born 1985, the year the original Katz/Shapiro/Farrell/Saloner/Arthur/David literature came out, and he still makes the same mistakes.
But. Going back to the article about Friston. This is a key point.
Random mutation is a key modeling ingredient of the strategic interaction & learning literature going back all the way to Kandori Mailath & Rob in 1993. Mutation is a key driver to reach "good" equilibria, but the arguments why they were included in the model amount to ¯\_(ツ)_/¯
This includes my own paper.
Handing the microphone back over to Karl Friston.
This just coming in via @delong. Note the error: increasing and decreasing returns are not juxtapositions of each other. They are two phases of a tatonnement process to find an efficiency frontier. Hype-disappointment cycles are a type of inc-dec returns.
But how do we get from wood lice to human behavior, which is hopefully much more purposeful than random straggling in a desperate attempt to get out of the searing sun? Well, to a certain degree. (Note also how much the article dwells on Friston's own oddball habits.)
Indeed, there is a whole literature about the unique psychological mindset of "the entrepreneur" going back to Schumpeter, there's no little consternation about how this erratic behavior doesn't stop when entrepreneurs become successful...
There's still a big leap to make on whether the core claim, that the kind of behavior Friston observes in his patients, and "entrepreneurial society" as a highly erratic, collective search function which can get stuck in temporary equilibria are connected by the same equations...
"Free energy" as human objective function has the huge advantage that it expressly doesn't work from rational behavior assumptions, but allows for a modicum of irrationality. Indeed it also explains artistic expression as a useful activity and not just societal slack.
And unlike the article, which gets hung up on the "fiendishly difficult math" (I have more to say about that), the objective function of the Boltzmann machine is actually fairly compact, and can easily be recalibrated to represent multiple game types.
Most of the coordination game literature starts from a specific type of (2-player) game and scales from there. Turn out they're closely related. Some of them don't even look like coordination games at all.
So life is indeed a Boltzmann machine. And unlike the rational actor model, which (as even the Chicago School admits) is a counterfactual model imposed on human decision making, "free energy minimization" might actually match up to what we observe on the synaptic level.
This wraps up my roadmap for economics for the next ~20 years.

I might have some footnotes...
FN1. Here's one for @VitalikButerin. Equilibrium convergence as a stochastic state machine (zero mutation in this case, just random sequence of moves). I actually had the hardest time explaining to IO economists what an automaton is, that's why I renamed it to "game graph".
FN2. I recommend that everyone read Beatrice's (@Undercoverhist) amazing history of economics vs engineering at Stanford. This topic has been the place where the tectonic plates of engineering, economics, physics, cognitive science have clashed for years.
FN3. Why I know that this will be the roadmap for economics for the next 20 years: Because I've seen a lot of people on all sides try to hammer away to make the tectonic plates match up, and I've seen many people come close (I consider myself among them).
FN4. The maddeningly complicated math contains some really exciting stuff. Markov blankets and Bayesian networks are indeed useful to model on the cognitive level. For economics it's more important to get the math down to an explanatory level.
In this case the difference between collective behavior (the "harmony" function H) and welfare-optimal behavior (W). To figure out the meaning and consequences of the "1/2" took me a few years.
Simplified it means "the only way we could reach a socially optimal outcome is if everyone acted considerate* all the time. But selfish behavior is a good enough approximation that the difference might not become apparent for a while."

*Take everyone else's utility into account.
Missing some Tweet in this thread?
You can try to force a refresh.

Like this thread? Get email updates or save it to PDF!

Subscribe to oliver beige
Profile picture

Get real-time email alerts when new unrolls are available from this author!

This content may be removed anytime!

Twitter may remove this content at anytime, convert it as a PDF, save and print for later use!

Try unrolling a thread yourself!

how to unroll video

1) Follow Thread Reader App on Twitter so you can easily mention us!

2) Go to a Twitter thread (series of Tweets by the same owner) and mention us with a keyword "unroll" @threadreaderapp unroll

You can practice here first or read more on our help page!

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just three indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member and get exclusive features!

Premium member ($30.00/year)

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!