So I’ve been toying with a philosophical framework that explains SourceCred to me: that it’s a supercharged hive mind, organized by voting. Each or interaction a vote in a mini election. The outcome of this swarm of elections continually shaping the graph. Perhaps this is why, when communities first turn on SourceCred, it often creates aha! moments. Perhaps it’s because the overall scores reflect people’s fuzzy, unexpressed reality. Or, perhaps their political reality, for better or worse. So when I recently came across a promising new (for me) theory in neuroscience, which reflects this idea and is being explored to to create more intelligent AI, my neurons started buzzing.
The Thousand Brains theory speculates that the brain is composed not of specialized circuits, organized hierarchically to process information for different tasks (the dominant view), but is instead composed of thousands of independent ‘brains’, each very similar in structure, continuously ‘voting’ on what is reality From article
Core to the theory is the surprising notion that the brain does not contain one model of the world; it contains thousands of complementary models for everything we know. The models vote together to produce our singular perception…there is not one such model but thousands, one in each of the many neatly stacked columns that constitute the brain’s cortex…cortical columns, in their world-modeling activities, work semi-autonomously. What we perceive is a kind of democratic consensus from among them. Democracy in the brain? Consensus, and even dispute? What an amazing idea.
One interesting conclusion is that the brain has some hierarchy, but much less than typically imagined. Each mini ‘brain’ effectively has the same basic circuits, like a general purpose computer processor. But is specialized somewhat at ‘birth’ by the area on the cortex it grows, and then specializes further based on the type of data it processes (e.g. vision vs. hearing vs. abstract thoughts). Here’s a more technical talk the theory’s creator gave at Microsoft a couple years ago, if you want more details il (starts after academic’y intros).
Sound familiar? What if the SourceCred Cred scores are beginning to reflect a cohesive “meta brain”, created by processing units (contributors, bots and plugins) constantly voting on a series of models (contributions)? What if our unique views, skills and experience, shape our ‘mini brains’ to optimize for different types of models? But can be, like real brains, retrained given new input (i.e. brain plasticity).
Is it possible that SourceCred’s founders, coming from Google Brain, created, along with the SourceCred algorithm and community, a living meta mind shaped after their own? Is the SourceCred meta brain a multiple?
Online communities and social networks have long been called ‘hive minds’. However, they’ve not been terribly intelligent thus far (see Twitter electing Donald Trump for lolz). However, what if by layering a Cred graph on top of the social graph, directly flowing real value (rent money level) along it, starting small (decentralized), and giving community members the ability to vote on their models of reality, we have created something more cohesive? Something inherently decentralized, but still able to coordinate surprisingly well?
In critiques of capitalism, I keep hearing the claim that capitalism has removed the political economy from our system, stripping away all political (i.e. human) values from economic policy and decisions. Alternatively, some worry that we’re actually just automating the political economy, with the algorithms and their value judgements (there are always value judgements) created by our new tech overlords, who, optimizing for profit, exacerbate existing inequalities and create new forms of oppression.
With our continual tweaking of the algorithm and its parameters, and daily swarms of “mini elections” shaping the graph, are we not to an extent automating our political economy, vote by vote? Intentionally making algorithmic decisions to create the world we want to live in? I’d like to believe so.
Perhaps SourceCred is better in some ways, despite conflicts, because the automation is guided by community members’ inputs. As our ability in the traditional legal system to pass laws to address new technology lags further and further behind said new technology, maybe a little automation is what we need.
Or, I’m way off base and need to cut down on the SourcePills? Curious to hear any thoughts from the meta brain.