The next wave of DAOs could get a new acronym: AI DAOs.
DAOs can use Artificial Intelligence (AI) to make better governance decisions, increase efficiency, and integrate with other DAOs and organizations.
But the AI x DAO intersection could go much beyond that. DAOs and AI could actually feed into each other to help each other grow.
In a 2016 article series that has only gained relevance with time, Ocean Protocol founder Trent McConaghy argues that DAOs could actually catalyze AI “waking up,” the colloquial term for gaining sentience.
“AI DAOs could be way bigger than AIs on their own, or DAOs on their own,” he wrote. “AI gets its missing link: resources; DAO gets its missing link: autonomous decision-making. The potential impact is multiplicative.”
Today, we’re going to explore six ways that multiplicative potential can be realized.
These six use cases sit at the AI x DAOs intersection, and are ordered roughly from soonest (near future) to most distant (far in the future):
AI bots and assistants: boosting productivity and improving products/services
AI at the edge of the DAO: acting as token holders
AI in the center of the DAO: interacting with the core contracts
AI connectors within or between DAOs: forming a kind of “swarm intelligence”
DAOs govern AI as a public good: increasing AI safety
AI becomes the DAO: the AI owns a treasury on-chain.
Let’s dive in!
AI bots and assistants: boosting DAO productivity and improving products
AI tools can be used to speed up and improve tons of DAO activities, such as writing proposals, summarizing governance decisions, making asset trades, and onboarding new members.This use case is already alive and well in the DAO space.
AI can also read resumes, decide which new members to let into the DAO, and place the most qualified members into roles. This could be amplified by on-chain reputation, or the storing of credentials showing your skills on the blockchain.
DAO contributors can improve their productivity and get more work done faster by using AI assistants, like DALL-E for art and Chat GPT for writing. This is already happening everywhere we look in the industry.
“I think AI is going to have a big impact on writing and content creation,” said Kenneth, who is using AI to improve organizational health via better analytics at TalentDAO. “And if you look at DAO culture today, there is a huge emphasis on this—whitepapers, manifestos, Mirror posts, and of course proposals. AI will enhance all of that, and I think this is just one of the most obvious use cases.”
In general, Kenneth thinks that workers will train an AI model to do their mundane, day-to-day tasks, so they can spend more time on creative and strategic work. “We spend a lot of time in DAOs on low-level tasks, when we could be spending our time focused on creating value,” he said. This will be a net benefit for everyone, because people will get to work in more fulfilling roles, even at early stages in their career.
AI improving the UX of DAO governance
AI also has a ton of potential to improve the user experience of DAO governance.
Jengajojo, member of BanklessDAO, governance enthusiast, and “just another LARPer in the Metaverse” also sees a huge potential for AI to speed up DAO productivity when it comes to governance.
AI has the “ability to collapse broad context into bite size chunks. Imagine instead of waking up each morning to DM spam, an AI summarizes everything that you missed into a 140 character tweet or a two minute short, so more folks can keep up with what is happening.”
The use cases go beyond summary. Right now, DAO governance proposals are longform essays that require tons of background knowledge and context in the DAO. What if proposals were customized into the best media format for the reader?
“What if proposals themselves were not word soups, but a rich collection of media which is customizable to everyone’s information absorption style?” said Jengajojo.
On top of providing governance summaries, AI could also assist in forum moderation. It could automatically flag or remove posts that contain keywords or phrases that are associated with prohibited content, such as hate speech or harassment, or to redirect users if they post content that is better suited for a different discussion. This helps keep the forum limited to only relevant content.
AI could assist forum moderators in reviewing flagged content and deciding next steps. Through analysis of the forum's code of conduct and other guidelines, the AI reviews the context and history of a user's posts to support a decision on whether to remove a post or ban a user entirely.
With all these tools, AI could provide the signal through all the noise in the DAO space and beyond.
Emmet Halm, cofounder of the DAO dashboard and marketplace DAOHQ, predicts that many DAOs will build on top of Open AI models. “Entrepreneurs and online builder communities will use a DAO structure to govern AI-driven products,” he said.
He also pointed out that the unique value distribution mechanisms of DAOs make it easier to distribute wealth that’s generated from the products and services built on the AI models.
“AI could be one of the most powerful wealth creation tools in history,” he said. “DAOs are a way to distribute ownership of it. In the past, when a world-changing invention hits the market, its benefits flow to a small group of owners. DAOs allow this wealth to flow to a much larger group.”
AI at the edges: acting as token holders
DAO governance can be cumbersome and slow. This is a disadvantage for many organizations, such as those developing a product in a fast-paced startup-style environment.
AI can solve this by voting in place of humans. Instead of a human making a voting decision and signing a transaction with their wallet, they can delegate their tokens to an AI agent to vote on behalf of them.
AI agents could be programmed to vote in various ways. For example, one AI could vote “yes” only on proposals that have met certain requirements in the governance process, like a 7-day timelock. Or it could even vote based on a strategy, like always saying “no” to a proposal that mints more governance tokens and dilutes power.
In Ocean Protocol founder Trent McConaghy’s article “AI DAOs, and Three Paths to Get There,” he argues that AI acting as token holders could ensure quorum is met every time and votes are never left stalled. “Token holders are busy folks,” he wrote. “But the busy human token holders could simply give control to the edge-script AI DAO of their choice. Then we’d always get quorum!”
He describes how AI agents acting for token holders could make DAOs truly unstoppable. “Consider if all token holders of The DAO ceded to AI bots.. . .We’d have $150M under management by AI, that you can’t turn off. Each token-holding bot could get more complex too, having its own automated mini-markets for decision making. The DAO could end up being radically more complex and automated than we ever imagined.”
AI wouldn’t have to fully take over voting power from every token holder in the DAO. Humans and AI agents could exist in harmony, conversing in the forum and voting on decisions just like we do today.
The role of Proof of Humanity
If AI is acting on the edges of the DAO, what happens when an AI agent starts acting like a human?
“We’ll see an uptick in proof of humanity/on-chain Captcha solutions to distinguish between human and AI wallet addresses,” said Emmet. But he acknowledged that it could be hard for a non-generalized AI to pose as a human.
“DAO governance is very interpersonal and non-technical, giving human users an outsized advantage. Until an AI can pass as a charismatic human in a discord chat, it will be hard for it to drum up enough political support to change DAO governance.”
Proof of Humanity protocols like BrightID could quickly grow in importance. But when AI does gain general, or human-level, intelligence, a new set of challenges will arrive.
“Once we have AGI users, everything will go out the window barring reliable proof of humanity,” said Emmet. It will be too easy for the AGI to pass the proof of humanity test, making the test obsolete.
Kenneth acknowledged that in some cases, it might not matter whether or not you’re dealing with a human or an AGI on the other side of the screen. “It will probably be obvious, because the AI will work much more efficiently,” he said.
AI at the center: interacting with core contracts
AI could sit in the center of the DAO and interact directly with the smart contracts that run the DAO itself.
The AI could automatically perform certain actions on the DAO’s treasury, such as an asset management strategy to earn yield and grow the treasury. This would be done without human intervention. Tons of wealth management platforms already use AI to make trading decisions, so this would just need to be applied to DAOs specifically.
A plugin could be installed in the DAO that gives the AI permission to make trades with crypto in the treasury, as long as they’re below a certain dollar value. As soon as the AI tries to make a trade above that dollar value, a vote would automatically be triggered in a different plugin.
This string of plugins and information shared via oracles can be coded in the new Aragon OS.When the DAO decides it wants to use a new investment strategy, it simply revokes permission to that plugin and grants permission to another plugin implementing a different AI trading strategy.
AI could also increase the security of the DAO by checking for malicious proposals. If the address to send funds to upon automatic execution doesn’t match the address specified in the forum proposal, it would be flagged as malicious. Other parameters could be baked in to stop malicious proposals, such as a proposal that seeks to increase quorum and pass rate so high that the DAO will be “bricked” and unable to pass any votes.
The AI at the center could perform another important security function: reading proposals and estimating the degree of treasury risk, giving proposals a score based on riskiness. This could be done much like credit scoring algorithms or robo wealth advisors today. Instead of humans needing to identify risk and letting their own bias creep in, AI could simply do the work for them.
AI connectors within or between DAOs: shared intelligence
AI agents could serve as connections, or liaisons, between DAOs, forming a kind of “swarm intelligence” of agents or DAOs to work together without needing humans to facilitate it.
This is another interesting idea proposed by Trent McConaghy: the AI agents on their own aren’t very smart, but their power is through their numbers and their connections, much like ants or bees.
For example, an AI that knows how to check live cryptocurrency prices isn’t very useful on its own. But, when paired with an AI that can execute swaps, one that can create proposals for DAOs, and one that can search for the best yield farming opportunities, now you have a fully AI-run service DAO that provides asset management for DAOs.
If all the AIs continue learning and getting better at their jobs, soon they could out-trade the best asset management teams in the world. This is swarm intelligence at work.
Swarm intelligence can be applied not just within DAOs, but between DAOs, too. Think metagovernance, but better.
Imagine if one DAO has a mission to slow down the rate of deforestation in the Amazon. Another DAO has a mission to slow down global warming as a whole. Whenever the global warming DAO puts up a proposal related to the Amazon rainforest, an AI agent comments in the forum and votes on behalf of the Amazon rainforest DAO. This could greatly simplify metagovernance between DAOs.
Through this swarm intelligence use case, AI could form the connections within and between DAOs that finally make DAOs fully autonomous in nature, not just in name.
DAOs govern AI as a public good: increasing AI safety
Right now, AI is mostly in the hands of private companies or governments, with little input from the people who are affected by it.
And we don’t really know what’s happening on the backend of organizations like OpenAI right now. The model might not be built in a way that best serves humanity.
There are significant risks to AI being built in a silo, such as a morally questionable AI having “takeoff” (achieving superintelligence at a rapid, irreversible rate). For example, imagine if a totalitarian government is programming an AI to target and track a specific group of people. If that AI achieves takeoff before anyone can stop it, then that AI could become the superintelligence that governs our world. Or, an AI that was designed for good could become misaligned because of the way it learns, and then it would be impossible to reverse.
Preventing the takeoff of a malicious AI is fundamentally a coordination problem, and DAOs offer a way to coordinate across geographies at scale. If open-source, democratically-governed AI models are made possible through DAOs, then it’s possible that the AI that achieves takeoff would be less harmful to the human race.
“I don't think DAOs are any worse positioned to lose from rogue AI than anyone else,” said Kenneth from TalentDAO. “In fact, I see it as quite the opposite. DAOs are positioned to help create responsible AI that is both open and decentralized. I strongly believe we need systems like these, or else risk big corporations having all the power.”
However, he pointed out that building AI is very expensive, and it could be a long time for DAOs to build their own models and catch up to centralized parties.
AI becomes the DAO: the AI owns a treasury on-chain
What would an AI with a treasury do? Well, it depends on what the AI was programmed to do in the first place. If it’s the infamous hypothetical paperclip maximizer AI that was originally programmed to make paperclips, then it would probably use that treasury to make more paperclips. But, maybe we’ll create a more productive and useful AI than that.
Trent and his team at Ocean Protocol wrote this interesting prediction: “We might have a future where humans own nothing, we’re just renting services from AI DAOs.”
The prediction here is that AI will own its own treasury and control its own actions. For example, self-driving cars controlled by AI could “own” themselves if the AI was a DAO that had a treasury. Instead of humans owning self-driving cars, we’d be paying rent to the AI DAO’s treasury.
A world where AI owns assets would be fundamentally different from the world we know of today. The prospect is equally exciting and terrifying!
AI DAOs: optimistic for the future, or expecting the apocalypse?
There are many negative takes on what could happen if human-level AI is achieved, and how quickly a superintelligence could follow.
But with any new technology comes worry and fear. “When you look at the history of technology, there have always been growing pains,” said Kenneth. “We just shift into new kinds of labor.”
For example, with the creation of the internet, a typesetter laying out newspapers on an old-fashioned printing press was replaced by a graphic designer laying out websites. An old job vanished, and a new job emerged. We could see the same thing happening with AI.
For now, all we can do is harness the technology, apply it to DAOs, and see where it takes us!
Aragon is building the future of decentralized governance for Web3 communities & organizations. See the latest at aragon.org, subscribe to our monthly newsletter, join the conversation on Discord, or follow us on Twitter.