A prominent subject within the realm of digital consensus (a contemporary term for cryptocurrency 2.0 that I’m currently experimenting with) is the idea of decentralized autonomous entities. Numerous organizations are swiftly entering this space, such as Bitshares (also referred to as Invictus Innovations) focusing on creating “decentralized autonomous companies,” BitAngels’ David Johnston with decentralized applications, and our own ideology surrounding decentralized autonomous corporations, which have evolved into the broader and less necessarily financial concept of “decentralized autonomous organizations” (DAOs); altogether, it can be asserted that “DAOism” is on its path to achieving a quasi-cyber-religion status. Nevertheless, one concealed issue that resides within this domain is quite apparent: individuals do not even comprehend what all of these specific terms signify. Precisely what constitutes a decentralized organization, how does one differentiate between an organization and an application, and what inherently qualifies something as autonomous from the outset? Many of us have experienced frustration due to the absence of consistent terminology in this area; as Bitshares’ Daniel Larimer highlights, “everyone imagines a DAC is merely a method for IPOing your centralized enterprise.” The objective of this article will be to investigate some of these notions and explore whether we can establish at least the foundations of a coherent comprehension of what all of these components actually entail.
Smart contracts
A smart contract represents the most basic form of decentralized automation and can be most precisely described in the following manner: a smart contract is a system involving digital assets and two or more participants, where some or all of the participants contribute assets, and those assets are subsequently redistributed automatically among the participants according to a formula that relies on specific data which remains unknown when the contract is initiated.
An illustration of a smart contract could be an employment agreement: A intends to compensate
The essential feature of a smart contract is straightforward: there is a fixed number of participants. The participants do not need to be entirely identified at the time of initialization; a sell order, where A proposes to sell 50 units of asset A to anyone who can provide 10 units of asset B, also qualifies as a smart contract. Smart contracts have the capacity to operate indefinitely; examples such as hedging contracts and escrow contracts illustrate this point. However, smart contracts that run indefinitely must still maintain a fixed number of parties (e.g., an entire decentralized exchange does not constitute a smart contract),and agreements that aren’t meant to persist indefinitely are referred to as smart contracts, as lasting for a limited duration inherently involves a finite number of participants.
It is important to highlight a specific gray area: contracts that have a finite nature from one perspective but an infinite nature from another. For instance, if I aim to hedge the value of my digital belongings, I might establish a contract allowing anyone to freely join or exit. Thus, the alternative side of the contract, involving individuals speculating on the asset at 2x leverage, can involve unlimited participants, while my side does not. Therefore, I suggest the following distinction: if the side with a limited number of participants is the one seeking a particular service (i.e., acting as a consumer), then it is categorized as a smart contract; however, if the side with a limited number of participants is merely in it for financial gain (i.e., acting as a producer), then it is not.
Autonomous Agents
Autonomous agents represent the opposite end of the automation spectrum; in such agents, there is no essential specific human involvement whatsoever; in other words, while some level of human effort might be required to construct the hardware on which the agent operates, there is no necessity for any humans to be aware of the agent’s existence. One illustration of an autonomous agent that already exists today is a computer virus; the virus thrives by replicating itself from one machine to another without intentional human intervention, and exists similarly to a biological organism. A more benign example would be a decentralized self-replicating cloud computing service; this type of system would initially operate an automated business on a single virtual private server, and as its profits grow, it would lease additional servers and install its own software on those, integrating them into its network.
A fully autonomous agent, or complete artificial intelligence, is the aspiration of science fiction; such an entity would be capable of adapting to arbitrary changes in circumstances and, theoretically, even expand to produce the hardware necessary for its own sustainability. Between a self-replicating cloud service and singular purpose agents like computer viruses lies a vast array of possibilities, which can alternatively be characterized as intelligence or adaptability. For instance, the self-replicating cloud service, in its most basic form, would only be able to rent servers from a limited selection of providers (e.g., Amazon, Microtronix, and Namecheap). However, a more intricate iteration should be able to determine how to rent a server from any provider by simply having a link to its website and utilize any search engine to discover new sites (and consequently, new search engines in case Google becomes unavailable). The next level would encompass upgrading its software, perhaps employing evolutionary algorithms, or adapting to new paradigms of server leasing (e.g., offering ordinary users an incentive to install its software and earn revenue from their desktops), and then the penultimate stage involves discovering and entering new industries (the ultimate stage, of course, is fully generalizing into an advanced AI).
Creating autonomous agents is among the most challenging endeavors, as to be effective they must be able to navigate an environment that is not only complex and rapidly changing, but also hostile. If a web hosting service intends to act unscrupulously, they could specifically target all instances of the service and then replace them with nodes that exploit the system in some manner; an autonomous agent must possess the capability to identify such exploitation and eliminate or at least neutralize deceitful nodes from the network.
Decentralized Applications
A decentralized application resembles a smart contract but varies in two important aspects. Firstly, a decentralized application features an unlimited number of participants on all sides of the marketplace. Secondly, a decentralized application does not necessarily have to be financial in nature. Due to this second criterion, decentralized applications are often among the simplest to develop (or at least, were the simplest before the emergence of generalized digital consensus platforms). For instance, BitTorrent qualifies as a decentralized application, along with Popcorn Time, BitMessage, Tor, and Maidsafe (it is noteworthy that Maidsafe is also a platform for other decentralized applications).
In general, decentralized applications can be categorized into two classes, likely with a significant gray area in between. The first class includes fully anonymous decentralized applications. In this case, the identity of the nodes is irrelevant; every participant remains essentially anonymous, and the system comprises a series of instantaneous atomic interactions. Examples include BitTorrent and BitMessage. The second class is a reputation-based decentralized application, in which the system (or at least certain nodes within the system) monitors the nodes, and those nodes uphold their status within the application through a mechanism that is solely maintained to foster trust. Status should not be transferable or possess de-facto monetary value. Maidsafe exemplifies this category. Naturally, absolute purity is unattainable – even a BitTorrent-like system must ensure peers maintain reputation-like statistics about other peers for anti-DDoS measures; however, the role that these statistics play is solely in the background and quite limited in scope.
A fascinating gray area between decentralized applications and “something else” pertains to applications like Bitcoin and Namecoin; these deviate from traditional applications as they create ecosystems and involve a concept of virtual property that holds value within this ecosystem’s context, such as bitcoins in Bitcoin’s case and namecoins and domain names in Namecoin’s case. As discussed below, my classification of decentralized autonomous organizations intersects with such notions, and there is ambiguity regarding their exact categorization.
Decentralized Organizations
In essence, a human organization can be defined as a combination of two components: a collection of property, and a protocol for a group of individuals, which may or may not be segmented into certain categories with varying conditions for entering or exiting the group, facilitating interaction among themselves, including rules for under what situations individuals may utilize specific portions of the property. For instance, consider a straightforward corporation operating a chain of stores. This corporation comprises three categories of members: investors, employees, and customers. The membership criterion for investors involves a fixed-size (or optionally quorum-adjustable size) portion of virtual property; one acquires some virtual property to join, becoming an investor until their shares are sold. Employees must be hired by either investors or other employees specifically sanctioned by investors (or employees authorized by other employees sanctioned by investors, and so forth recursively) to engage, and can also be dismissed in a similar manner, while customers represent an open-membership structure where anyone can freely interact with the store in the officially recognized manner at any time. In this model, suppliers are akin to employees. A nonprofit charity varies in structure somewhat, involving donors and members (charity recipients may or may not be regarded as members; the alternative perspective considers the positive advancements in the recipients’ well-being as the charity’s “product”).
The concept of a decentralized organization adopts the same principle of anorganization, and disperses it. Rather than a vertical framework overseen by a group of individuals interacting face-to-face and managing assets through the legal framework, a decentralized organization consists of a group of individuals engaging with one another based on a protocol defined in code, and upheld on the blockchain. A DO might utilize the legal system for some defense of its tangible assets, but even in those instances, such usage is subordinate. For illustrative purposes, one could take the corporation owned by shareholders mentioned earlier, and transition it entirely onto the blockchain; a long-lasting blockchain-based agreement preserves a record of each person’s shareholdings, and voting on the blockchain would permit shareholders to choose the board of directors and the staff. Systems for smart property could also be directly integrated into the blockchain, potentially enabling DOs to manage vehicles, safety deposit boxes, and buildings.
Decentralized Autonomous Organizations
In this section, we delve into what may be seen as the ultimate goal, the concept that possesses the most ambiguous definition of all: decentralized autonomous organizations, along with their corporate subclass, decentralized autonomous corporations (or, more recently, “companies”). The vision of a decentralized autonomous organization is straightforward to articulate: it is an entity that resides on the internet and exists independently, but also heavily depends on recruiting individuals to execute specific tasks that the automaton itself cannot accomplish.
Considering the above, the crucial aspect of the definition is really to emphasize what a DAO is not, and what does not qualify as a DAO but is instead a DO, a DA, or an automated agent/AI. To begin with, let’s examine DAs. The primary distinction between a DA and a DAO is that a DAO has internal capital; that is, a DAO possesses some form of internal asset that holds value in some respect, and it has the capability to utilize that asset as a method for rewarding certain activities. BitTorrent lacks internal assets, while Bitcloud/Maidsafe-like systems may have reputation but that reputation is not an exchangeable commodity. Bitcoin and Namecoin, conversely, do possess internal capital. However, traditional DOs also have internal assets, similarly to autonomous agents.
Next, let’s examine DOs. The evident difference between a DO and a DAO, inherently implied in the terminology, is the term “autonomous”; that is, in a DO, the individuals are the ones making the decisions, whereas a DAO is something that, in some manner, makes decisions independently. This is a surprisingly complex distinction to clarify because, as dictatorships often emphasize, there is essentially no difference between a certain group of individuals making decisions directly and that group managing all of the information upon which decisions are based. In Bitcoin, a 51% attack among a limited number of mining pools can cause the blockchain to reverse transactions, and in a hypothetical decentralized autonomous corporation, the data providers can collude to make the DAC believe that sending all of its funds to1FxkfJQLJTXpW6QmxGT6oF43ZH959ns8Cq is equivalent to paying for a million nodes’ worth of computing power for a decade. However, there is undeniably a significant distinction between the two, and thus we need to articulate it.
My own attempt at defining the difference is as follows. DOs and DAOs are both susceptible to collusion attacks, where (in the best-case scenario) a majority or (in worse scenarios) a substantial percentage of a specific type of members collude to steer the D*O’s activities. However, the distinction is this: in a DAO, collusion attacks are regarded as a flaw, whereas in a DO, they are a feature. In a democracy, for instance, the fundamental purpose is that a plurality of members selects what they prefer most and that solution is implemented; in Bitcoin, however, the “default” behavior that occurs when everyone acts according to their individual interests without a predetermined outcome is the intention, and a 51% attack favoring a specific blockchain is an anomaly. This appeal to social consensus is akin to the definition of a government: if a local gang begins imposing a property tax on all shopkeepers, it may evade repercussions in certain regions, yet no significant segment of the population will consider it legitimate, whereas if a government were to adopt the same practice, the public reaction would likely differ.
Bitcoin presents an intriguing case in this respect. Generally, it appears to be much more akin to a DAO than a DO. However, there was one occurrence in 2013 where the reality turned out to be notably different. What transpired was that an unusual block was (at least we hope) unintentionally constructed, which was deemed valid according to the BitcoinQt 0.8 clients but invalid according to the standards of BitcoinQt 0.7. The blockchain forked, with some nodes adhering to the chain following this exceptional block (which we’ll refer to as chain B1), while the other nodes that regarded that block as invalid operated on an alternate blockchain (which we’ll call B2). Most mining pools had upgraded to BitcoinQt 0.8 and thus followed B1, but the majority of users remained on 0.7, following B2. The mining pool operators convened on IRC chat and decided to shift their pools to mine on B2, since this outcome would simplify the experience for users by not requiring them to upgrade. After six hours, chain B2 surpassed B1 due to this intentional action, leading to B1 being abandoned. Consequently, in this instance, there was a deliberate 51% attack that the community accepted as legitimate, positioning Bitcoin as a DO rather than a DAO. Nonetheless, in most instances, this scenario does not occur, so the most appropriate classification for Bitcoin would be as a DAO with an imperfection in its autonomy implementation.
Nevertheless, some are not satisfied with categorizing Bitcoin as a DAO, arguing that it lacks sufficient intelligence. Bitcoin lacks thought processes, does not proactively “hire” individuals, except within the mining protocol, and adheres to straightforward rules where the upgrading procedure resembles more of a DO than a DAO. Advocates of this perspective would define a DAO as possessing a considerable degree of autonomous intelligence. However, the problem with this outlook is that a distinction must be drawn between a DAO and an AA/AI. The differentiation here can be characterized as such: an AI operates completely autonomously, whereas a DAO still necessitates substantial engagement from humans specifically collaborating according to a protocol established by the DAO for its functionality. We can categorize DAOs, DOs (and traditional Os), AIs, and a fourth category, traditional robots, using a classic quadrant chart, along with another quadrant chart to classify entities that do not possess internal capital, thereby creating a cube:
DAOs == automation at the core, humans on the outskirts. Consequently, overall, it makes the most sense to perceive Bitcoin and Namecoin as DAOs, albeit ones that barely surpass the DA threshold. Another crucial distinction is internal capital; a DAO devoid of internal capital is a DA, while an organization lacking internal capital is a forum; the G8, for instance, would fit the definition of a forum. DCs in the preceding graph refer to “decentralized communities”; an example of this might be something akin to a decentralized Reddit,where there exists a decentralized system, yet there is also a community surrounding that system, and it remains somewhat unclear whether the community or the protocol is genuinely “in charge”.
Decentralized Autonomous Corporations
Decentralized autonomous corporations/companies represent a more specialized subject, as they effectively constitute a subset of DAOs, although they merit attention. Since Daniel Larimer is the primary proponent of DAC as a term, we will adopt his definition that he repeatedly advocates: a DAC distributes dividends. In other words, there exists a concept of shares within a DAC that can be purchased and traded in various ways, and those shares may grant their owners continuous returns based on the DAC’s performance. A DAO operates without profit; although one can generate income within a DAO, this occurs through engagement in its ecosystem rather than by investing in the DAO itself. Clearly, this differentiation is somewhat ambiguous; all DAOs possess internal capital that can be owned, and the valuation of this internal capital can swiftly increase as the DAO gains strength/popularity, leading a significant portion of DAOs to resemble DACs to some degree.
Therefore, the distinction is more of a fluid one and depends on emphasis: to what degree are dividends the primary focus, and to what degree is it about earning tokens through participation? Furthermore, to what degree does the notion of a “share” exist as opposed to mere virtual property? For instance, membership on a nonprofit board is not truly a share, as membership is often granted and revoked at will, an issue unacceptable for something categorized as investable property, whereas a bitcoin does not represent a share due to not conferring any entitlement to profits or decision-making rights within the system, in contrast to a share in a corporation, which clearly qualifies as a share. Ultimately, the distinction might hinge on the surprisingly obscure question of whether the profit mechanism and the consensus mechanism are identical.
The aforementioned definitions remain far from exhaustive; there will likely be ambiguous areas and gaps within them, and determining the precise degree of automation a DO requires before it qualifies as a DAO is a challenging question to resolve. Additionally, there is also the matter of how all of these elements should be constructed. An AI, for instance, would probably function best as a network of private servers, each operating often proprietary local software, whereas a DO ought to be fully open-source and blockchain-based. At the midpoint of these two extremes lies a multitude of different frameworks to explore. How much intelligence ought to reside in the core code? Should genetic algorithms be employed for updating software, or should it be futarchy or some voting or vetting system based on individual input? Should membership be structured corporately, with sellable and transferable shares, or nonprofit-style, where members can vote to admit or remove others? Should blockchains follow proof of work, proof of stake, or reputation-based models? Should DAOs aim to sustain balances in alternative currencies, or should they only incentivize behavior by issuing their own internal tokens? These present complex challenges, and we have merely begun to scratch the surface of them.
