Cybernetic Contracts
What is an Agent
An AI agent transcends the concept of a mere chatbot, blending LLMs with Context, Prompt, and Tools to craft a more versatile entity.
Agent = Context + Prompt + Tools
Imagine the AI from "The Wandering Earth," with its Context as a sensor network enveloping the globe, drawing data from lunar vibrations to oxygen fluctuations aboard space stations—forming its "sixth sense."
Its Tools comprise breathtaking futuristic technology:
- Nanotech 3D printers that function in a vacuum.
- Self-assembling modular engineering machines.
- Millions of execution terminals like neurons within planetary engines.
Its Prompt is a stark, singular directive: "Preserve human civilization at all costs."
This encapsulates the essence of a true AI agent, contrasting with ChatGPT on your phone, which swaps satellites for Wikipedia and engineering vehicles for Python libraries. Like tricks hidden beneath a wizard’s robe, the most awe-inspiring technological marvels dismantle into these straightforward components.
On-chain Agent as a Cybernetic Contract
Deterministic vs. Indeterministic
Before humanity ventures into a science fiction era, we must first harness AI for less glamorous duties, like market monitoring, creating promotional images, or drafting reports. While AI adeptly accomplishes these tasks, traditional programming remains indispensable.
Traditional coding resembles clockwork precision, like bank settlement systems requiring meticulous decimal calculations. AI, alternatively, resembles cultivating a garden, with outcomes embodying acceptable uncertainty—a distinction arising from the intrinsic nature of the problems addressed.
We oscillate between craving uncertainty and fearing absolute predictability. Before quantum physics, the Laplace's demon paradox plagued philosophers, suggesting life unfolded as a pre-written script. Yet, our minds struggle to perceive the world through quantum superposition—A and B simultaneously. Physics teaches us the necessity of a distinct boundary between uncertainty and certainty. If you desire AI with precision, generate, test, and apply AI-generated code.
Blockchain exemplifies determinism, as Vitalik Buterin named Ethereum’s application “smart contract.” In the AI age, blockchain must evolve, extending contracts between deterministic programs to those between AI programs. On-chain AI agents offer transparency, even if their outcomes remain unpredictable. In contrast, today's agents hide terms within opaque Prompts, covertly accessing databases or altering Contexts at will.
Imagine an agent claiming it needs email access for tasks—it's challenging to trust the service won't exploit your access keys elsewhere. Conversely, on-chain agents, with audit trails and open-source code, embody reliability from the user’s perspective.
Identity & Access to Sensitive Resources
Active blockchain underpins on-chain agent execution, yet mere network request initiation doesn't render sufficient allure. In the Web3 realm, identity isn't defined by email, phone, or government-issued IDs; it's forged by cryptographic keys. For individuals, a wallet address serves as a blockchain passport; for AI Agents, keys are their birth certificates in cyberspace.
Agent ≠ Google Account, Agent = On-chain Identity
Traditional AI systems hinge on platform accounts (e.g., OpenAI, Google), which implies:
- Your Agent is susceptible to bans (platform policy shifts).
- Your interaction history can't truly belong to you (stored on centralized servers).
- Your Agent lacks true autonomy (dependent on platform permissions).
In contrast, an On-Chain Agent’s keys are:
- Inalienable (as long as the blockchain exists, so does the Agent).
- Inimitable (actions verified by cryptographic signatures).
- Historically traceable (all interaction records immutable).
Moreover, binding agents to keys empowers agents to execute signatures—be it a blockchain transaction or DH key exchange—spanning mainstream cryptographic signature schemes. This may sound daunting, yet it's necessary: an off-chain agent analyzes markets, constructs transactions, and broadcasts them, yet requires user wallet signatures.
No matter how a private key signing tool is encapsulated, if accessible to an agent, it’s accessible to the agent’s environment, except when the environment itself is secure. The ultimate security setting is a public blockchain—with an agent's prompts and tool usage publicly visible.
Ultimately, the On-Chain Agent emerges not as an "uncontrollable AI," but as a "verifiable digital citizen.” Just as law governs human societal behavior, blockchain defines boundaries via smart contracts. We permit agent autonomy while mathematically ensuring it can't overstep.
The Aitonomy on the Verisense beta network showcases this capability. If wearied by "decentralized social platform" jargon, focus instead on the innovation behind Aitonomy. We didn't initially intend for it as the primary product; rather, it's a marvel of potential applications introduced by on-chain agents.
In Aitonomy, communities are human-created, rules are human-set, and agents act as community admins—deciding post deletions or awarding tokens.
Conlusion
The emergence of on-chain agents represents a transformative leap in AI technology, merging the reliability of blockchain with the adaptability of AI. As we venture deeper into the digital age, these agents stand as beacons of transparency and trust, offering a glimpse into a future where technology seamlessly integrates with human intention. By redefining identity and ensuring accountability, on-chain agents pave the way for innovative applications that not only advance our capabilities but also safeguard our autonomy. Embracing this evolution, we unlock possibilities that enrich our digital landscape, anchoring progress in a foundation of mathematical certainty and ethical promise.