As Our Progress toward AGI Accelerates, So Must Our Progress on Decentralized Governance

The core mission of SingularityNET is to catalyze the emergence of beneficial Artificial General Intelligence. One of the main beliefs motivating the path SingularityNET has taken in pursuing this mission is that the “beneficial” part is more likely to be realized if AGI is decentralized in its underpinnings, rather than if it’s owned and controlled by a small elite group or a particular national, corporate or other sectarian interest.

But what actually does “decentralized” mean? It’s not just about software and networking. Yes, decentralized AGI requires sophisticated, scalable distributed processing, via algorithms enabling efficient use of widespread computing resources in architectures not requiring centralized controllers. And yes, this involves a variety of difficult computer science and software design problems, which are being addressed within the blockchain world as well as in other parts of the research community and software industry. The SingularityNET platform, the Hypercycle ledgerless blockchain, the NuNet decentralized compute resource allocation protocol and other SingularityNET network innovations are all contributing here significantly. But until the AGIs are really running the show without human supervision, even the best decentralized software in the world won’t manifest itself as a decentralized AI network unless the network of humans underlying the software and hardware involved is also organized in an appropriately decentralized manner.

As a well-worn (but unfortunately very realistic) example, a blockchain network governed by a “one token, one vote” mechanism is not actually terribly decentralized if a majority of tokens and therefore votes are held by a small number of parties.

The SingularityNET plan from the get-go was to get the ball rolling with a small group of enthusiasts building software and spreading the vision, and then over time transition to a more fully decentralized mode of organization in which design and development decisions are made by a broader, heterogeneous community, and overall governance takes the form of some species of participatory democracy. The reason for taking this sort of “progressive decentralization” approach is clear: Encouraging an active, productive community to form around a complex new idea with a mix of technical and nontechnical aspects is one thing; encouraging one to form around a working software system with a community of users is another. The latter is much more tractably achievable.

So, here we are. SingularityNET has by no means achieved all its goals set out at its founding in 2017, but we have a working core protocol serving as the back-end for a variety of AI tools (e.g. a well-liked song splitter soon to be followed by a number of other AI/music tools, and the Rejuve longevity app, among others) and a variety of additional major technical pieces beyond what it was originally realized would be needed, such as HyperCycle and NuNet, the AI-DSL and MeTTa as a smart contract language. We have a number of ecosystem projects working on practical domain-specific applications leveraging these tools, with varying levels of maturity (SingularityDAO, Rejuve, Sophiaverse, Mindplex, TWIN Protocol, TrueAGI, Zarqa, etc.). We have a new scalable AGI framework, OpenCog Hyperon, advancing steadily and slated for alpha release in early 2024, with a clear path to integration into the ecosystem’s overall decentralized toolset.

While no one can predict the precise details of technology development, especially in rapidly changing domains like AI and blockchain, it would appear that the various technologies under development in the SingularityNET ecosystem are poised for a remarkable convergence over the next 1–3 years. There is a real prospect of launching an AGI network combining OpenCog Hyperon nodes with nodes running LLMs and other AI algorithms, solving a variety of application problems for diverse ecosystem projects, and operating in a decentralized manner according to the SingularityNET ecosystem tools. The AI field has been accelerating recently in a truly dramatic way, and while much of the recent activity has been in the specific area of LLMs, it seems likely that the next phase of acceleration will involve hybrid systems integration LLMs with other forms of AI, and both Hyperon and the SingularityNET framework are ideal for enabling this sort of hybridization at scale.

If this projection holds any reality (and obviously I believe it does!), then it follows that RIGHT NOW is the time we must accelerate the process of progressive decentralization, to match the foreseen acceleration of the software of decentralized AGI. To realize the decentralized beneficial AGI vision that has underlain SingularityNET since its founding, we need a richly functional participatory-democracy governance fabric to already be in place and robustly operating by the time something resembling human-level general intelligence emerges on the SingularityNET network. And for those curious about the distinction between the powerful and exciting capabilities displayed by ChatGPT and similar systems and full-on human-capable AGI, please see here.

Once the first true breakthrough to human-capable AGI is made, a great variety of subtle decisions will present themselves, some needing to be made in a rapid way. Major governments and corporations will take more than a casual interest. Decentralization is not a magic bullet and doesn’t eliminate all risks associated with the advent of advanced AGI — but just as with the decentralized structures and dynamics underlying the Linux operating system and the Internet itself, we place more faith in global participatory modes of organization than in elites and other narrow interests, or self-appointed patriarchal guardians. If decentralized deployment, development and governance is the reality when advanced AGI emerges, this will be the reality that governments, corporations and humanity as a whole has to work with, and we believe this has higher odds of coming out well for humanity and other sentient beings than any available alternative.

Which brings us to the current Supervisory Council election. Different from the prior SingularityNET Supervisory Councils, the newly elected council will have a 6-month term, and a highly specific goal: To create a proposal, to be submitted to the AGIX token-holders for voting, regarding a concrete path from the current governance and organizational structure of SingularityNET Foundation toward a much more radically decentralized network with participatory governance woven in at every level.

Such a proposal is not expected to emerge from the new Supervisory Council alone — we who work in the Foundation have no shortage of ideas about decentralized governance, and the SingularityNET community is full of creative-minded individuals with enthusiasm to contribute to shaping the path to beneficial decentralized AGI. However, the role of the Supervisory Council in gathering, reconciling and synthesizing inputs from across the ecosystem into a coherent overall proposal will be a significant and challenging one.

The technical work underway in the SingularityNET ecosystem is among the most complex and subtle projects occurring anywhere on the planet today — it may sound a bit hubristic to make a statement like this, but the combination of computer science, math, software engineering, cognitive science, linguistics, and other advancement needed to create a beneficial decentralized AGI system really is remarkably intense. Which is part of what makes it fun to work on, of course, and part of why we have been able to attract such an extraordinarily skilled and visionary technical team. While not everything in the SingularityNET project has gone according to plan so far, we have succeeded in making tremendous technical progress which is gaining more and more momentum; from this standpoint, pursuing a major change of governance and organization could be perceived as a bit foolhardy. “If it ain’t broke don’t fix it” goes the saying, and in some very critical respects what we have in SingularityNET’s technical teams is a very well-functioning machine.

Ultimately to achieve its vision SingularityNET’s network will need to have the capability to function fully without human input — although appropriately ethical and effective operation will always involve having humans fully in the loop where human affairs are concerned. But until we pass through and beyond the human level of capability, the human governance and coordination aspect of SingularityNET is going to be every bit as important as the technical aspect. To achieve our goal of beneficial decentralized AGI we need an extremely well-functioning decentralized network of humans working together with our network of AI agents. With this in mind, let all of us who care about SingularityNET’s mission do what we can to form the new Supervisory Council in an optimal way and help it carry out its mission!

How to Join the Supervisory Council Election

Applications for candidacy may be submitted until Sunday 1st October and voting will take place between Wednesday 11th and Tuesday 17th October (inclusive).

This is a unique opportunity to contribute to SingularityNET’s vision to decentralize, support the community, and tackle persistent challenges in decentralization.

Please read the role description and eligibility requirements.

To apply, visit the Swae portal here, and be as detailed as you can about why you should be part of SNET’s second Supervisory Council!