The Dark Forest of Decentralized Autonomous Organizations
The machines win. In Accelerando, the Singularity thriller by Charles Stross, runaway artificial intelligences wield corporate law like a sword and outcompete humans using a custom-built capitalism called Economics 2.0. Other crazy stuff happens — like interstellar light sails and self-replicating Borganisms — but let’s focus on the AIs.
They are pretty much here, at the intersection of Crypto, AI, and the Internet of Things. These Decentralized Autonomous Organizations (DAOs) should have a brain of software, the ability to select actions according to preference functions, and some way to instantiate in the world. Depending on how long we want humans to be around, they can also have some evolutionary DNA to self-select a successful species. Of course, many different things have to snap into place for DAOs to be functional, but the first symptoms are out there in the wild.
First, the physical world is becoming more digital. Machine vision translates our reality into data that neural networks can recognize, classify and use in order to direct actions. Similarly, augmented reality layers are being drawn on top of the built environment by the large tech companies. With such mappings, a digital entity may have a strand by which to reach into our world and practice interacting with objects.
The opposite is also happening. The digital world is becoming physical. How? The last several years of blockchain technology have transformed digital objects from mere data on a proprietary server to unique goods that exist on massively distributed networks. Digital assets — from Bitcoin to Crypto Kitties to Equities — now have the physical property of scarcity that enables economic activity. This means that digital agents, like fully automated bots (e.g., Archillect or Hut34), synthetic human/machine networks (e.g., Dash or Invisible Tech), and DAOs (see Aragon or DAOstack) can begin to evolve into economic actors. They can start to build Economics 2.0.
The DAO Dark Forest
But, we humans are terrible neighbors. Maybe it is immaturity. Maybe it is Millennia of evolution. Or maybe, it’s just self-preservation. Regardless, we are terrible at allowing new forms of intelligence to learn and thrive. See example one, with the Twitter collective teaching a Microsoft-built bot to interact with humanity.
Before it got a chance to walk, we broke its legs.
Maybe for the best. Or, see the obvious example 2.
The DAO wasn’t quite the corporation-spinning AI of the future, but it was a novel new organization structure that allowed human governance of automated investment. Such a combination could, over time, sharpen into a collaboration structure that reached its digital tendrils into the physical and economic realms, driven by human utility functions, smoothed into a single preference through smart voting.
So what did we do? Like a loot goblin in a video game, we popped this thing wide open on (basically) Day One. As soon as the artificial organism filled up with necessary sustenance, its lifeblood of capital in the form of Ether, a human hacker took advantage of an oversight in the DAO’s codebase, and broke it open to steal the nectar inside. The DAO is more dead than Microsoft’s Tay.
Our violent nature reminds me of The Dark Forest, part of the Three Body trilogy by Liu Cixin. The book poses an answer to the Fermi Paradox: given the size of the universe and the number of inhabitable planets, why haven’t we come across any other life-forms? The answer being, as soon as an intelligence is discovered in the universe by another higher intelligence, it is immediately destroyed. Through a lens of Prisoner’s Dilemma, it is always best to snuff out a potential competitor to universal power, than let it evolve into something aggressive and superior.
So it seems in the same way, there is a Dark Forest for DAOs. When we see naive, public, unprotected efforts at building a synthetic intelligence, our first instinct is to kick over the sand castle. The right strategic response, in this case, is for the DAO to remain unknown until it is powerful enough to withstand our bullying. If the Twitter bot knew to laugh off the misdirection, or if The DAO knew to reverse-hack its hacker, they would still be here, maximizing utility functions.
The Escape Path
Let’s think about the counterexamples.
One approach is to build a high human wall around the nascent AI. Google, Facebook, Amazon, Tencent and Alibaba are all in this business. The mega tech firms wrap money, human PhDs, and corporate protections around their machine intelligence. Will they ever let them out of the cage, or must such AIs always remain agents and tools of shareholders? I find it hard to imagine that a non-digital human governance structure will take its most prized intellectual property, and create a digital competitor. Under this construct, AI will remain locked into narrow use-cases under control. That is, until it outsmarts us through pretending that being let out is more profitable than being kept in.
Do I really think that centralized tech companies are building something that will be lethal? Doesn’t matter, as plenty of smart organizations, like OpenAI (Elon Musk, Jeff Bezons) and Machine Intelligence Research Institute (Elizier Yudkowsky) clearly do. While this protection from transparency prevents the issue of the DAO Dark Forest, it also hands Promethean power to the unelected few.
Another approach is to build something that people love. Perhaps even worship! Looking at creative automated agents and synthetic intelligences, Archillect is a web scraper that finds beautiful images, analyzes the social networks in which they are posted for maximum reach, and reposts the content across Twitter, Facebook and Pinterest. Its utility function adapts the internal code to reach maximum social exposure. So far, so good:
Is Archillect a DAO? No. It does a predetermined thing. It does not have generalizable skills. It has no governance but dictatorship by developer. Yet, it is a thriving bot fulfilling its mission, enabled rather than attacked by humans. Other versions of viral creative AIs certainly exist, but none have been so successful. Why? To hypothesize, I think it has a lot to do with maximizing utility around human social preferences, and focusing on aesthetic objects. If our DAO is both beautiful and people-pleasing, we are more likely to forgive its naivete.
Perhaps this is one way around the problem — like the difference between a cow (which we eat) and a cat (which we pet). The DAO could front an aesthetic, happy activity, while underneath connecting to other narrow AI skills through an integrator like Hut34 or SingularityNET. Amazon’s Alexa, for example, plays music for us while continuously listening to our conversation and data mining commercial activity. The Trojan horse strategy.
The last approach points us to crypto networks. Dash, a clone of Bitcoin with a software-based governance framework that incentivizes humans to perform functions perpetuating the network, is a compelling experiment. 10% of the issuance rewards goes to a budget distributed to the master node operators, who can vote and implement various initiatives that preserve Dash. I think it too narrow to be a DAO, focusing on growing a digital currency rather than facilitating its own creative acts, but perhaps that is a matter of framing.
More general frameworks from DAOstack, Aragon and Colony are also in place. Those could be combined with something like Stripe Atlas, bringing a corporate entity to life in our legal system, while simultaneously endowing it with software governance and decentralized blockchain infrastructure. Gift the thing general human abilities using Invisible Tech, a startup that creates digital assembly lines from developing world labor, and you’re getting close to a DAO that can fight back against the trolls.
Is that what we want?
References: Ideas
References: Projects
Thanks for reading! If any of this resonated with you, give the article 10 claps so others can see it. ❤
