Since the "birth of Chatgpt" in 2023 and reaching the milestone of 100 million users in just two months, the field of artificial intelligence has become a very popular track for investment institutions. In the field of encryption, the combination of distributed systems, cryptography and other technologies with artificial intelligence is also very attractive to capital. In 2023 alone, the financing amount of the AI track in the Web3 industry was US$298 million, exceeding the NFT track.
Source: Rootdata, Binance Research, as of December 31, 2023
Related reading: "Binance Research: Latest Data and Development of AI+Crypto"
In November 2023, the decentralized AI computing platform Ritual announced the completion of a $25 million financing round, led by Archetype, with participation from Accomplice and Robot Ventures. It is reported that Ritual aims to create an incentive network to power distributed computing devices to support various applications of artificial intelligence. The funds will be used to build network infrastructure, expand the team, and develop the Ritual ecosystem.
“The consolidation of AI among a small group of powerful companies poses a significant threat to the future of technology. We founded Ritual to end the ecosystem’s dependence on a small number of people, open up access to this critical infrastructure, and ensure a better AI future. Ritual is the decentralized network that the ecosystem needs,” said Niraj Pant, co-founder of Ritual, in a statement.
AI and Crypto: A Two-Way Journey
Combining the best principles and techniques of cryptography and artificial intelligence, Ritual aims to create a system that enables the open and permissionless creation, distribution, and improvement of artificial intelligence models. Ritual seamlessly integrates artificial intelligence into any on-chain application or protocol, enabling users to fine-tune, monetize, and reason about models using cryptographic schemes.
Niraj Pant proposed five key focus areas for Ritual: creating an incentive network; connecting distributed computing devices to support hosting, sharing, inference, and refinement; adapting AI models to the API layer used to access models; a proof layer to ensure computational integrity; and resisting censorship and protecting privacy.
Niraj Pant said that currently the chips, computing power and models of AI tools are in the hands of a few companies, posing a threat to the future development of technology. Several core issues are:
Lack of strong SLAs: Existing platforms do not provide any guarantees about computational integrity (i.e., whether the model ran correctly), privacy (the inputs and outputs of the model), and censorship resistance (limiting scrutiny of the model, application, and geography);
Permissioned and centralized APIs: Existing infrastructure is hosted by a few centralized companies, limiting developers and users from building native integrations;
High computing costs and limited hardware access: It is becoming increasingly difficult for developers to procure AI hardware, and hardware providers charge developers high commissions;
Oligopoly and structural dysfunction: Organizations are either incentivized to keep their models closed, thereby inhibiting innovation and concentrating power, or to open source their models, recognizing the lack of appropriate infrastructure to reward their contributions. In addition, users have little say in the governance and ownership of AI today.
These problems can be solved using innovations in cryptography, game theory, and mechanism design. Ritual’s goal is to break the reliance on these companies and open up access to critical infrastructure to ensure better AI can be built.
AI technology can also bring new development momentum to the crypto field. From base layer infrastructure to applications, AI models can be used to encapsulate complex logic and enable new applications that were previously impossible with smart contracts alone. For example, we imagine a world where users can generate transactions and interact with contracts using natural language, or where agents automatically manage risk parameters for loan agreements based on real-time market conditions. There are a ton of fascinating use cases, but there is a lack of infrastructure that can bridge the gap between accessing models and leveraging them on-chain.
Ritual is at the forefront of this intersection and is building a unified solution to both problems.
Ritual
Ritual brings together a distributed network of nodes with access to computation and model creators, and enables all creators to host their models on these nodes. Users can then access any model on this network (whether it is an LLM or a classic ML model) using a common API, and the network has additional cryptographic infrastructure to guarantee computational integrity and privacy.
The components of Ritual are as follows:
Ritual Superchain: A set of sovereign modular execution layers, each containing specialized state precompilations (SPCs) suitable for different categories of arbitrary computations, mainly around AI models, from classic models to basic models. The GMP layer facilitates interoperability between existing blockchains and the Ritual Superchain, which acts as an AI coprocessor for all blockchains. Ritual's AI VM contains not only SPCs, but also base layer infrastructure that facilitates optimized execution, including inference engine binaries and vector databases.
Node Set: The Ritual Superchain consists of node categories, each containing different functionalities and resource requirements. Ritual nodes include canonical full nodes and validator nodes, as well as Ritual-specific nodes (including proof nodes, model cache nodes, and privacy nodes). Ritual proof and privacy nodes can leverage a variety of mechanisms, from ZK to Optimsim in terms of proofs, and FHE to MPC in terms of privacy, depending on the user's required guarantees and the complexity of the AI model.
About Stateful Precompiles (SPCs): Stateful precompiles are precompiles with state access. Ritual requires highly optimized operations that can efficiently compute specific functions of various AI models. Some SPCs can be implemented as a combination of other SPCs (i.e. fine-tuning and inference). Some SPCs can take advantage of various types of parallelism (i.e. embedding), while others are sequential by construction.
Generic Messaging Passing (GMP): Ritual enables applications on any chain to leverage the execution capabilities of the hyperchain through a compact, bidirectional generic messaging transport.
Portals: Portals are a unique feature of Ritual that allow for eager data evaluation on the source chain via native smart contracts before leveraging the Ritual superchain. Portals are optimized for static analysis of AI model inputs, localizing computation to the source chain, minimizing data sent over the network.
Hell
Infernet is the first building block in the suite of protocols and utilities that Ritual will release, and Infernet Node v0.1.0 is a lightweight off-chain client for Infernet serving computational workloads.
Infernet enables anyone to seamlessly build on top of Ritual and gain permissionless access to Ritual’s network of models and compute providers. Infernet brings AI to today’s on-chain applications by providing a powerful interface for smart contracts to access AI models for inference.
Ritual hopes to develop Infernet into a modular execution layer suite that interoperates with other base layer infrastructure in the ecosystem, becoming a key point for AI in the web3 space, allowing every protocol and application on any chain to use the Ritual AI coprocessor. Using a given model and function, a smart contract can request Infernet to calculate some outputs and proofs.
The Infernet workflow can be understood through Frenrug, an interactive experimental example provided by the Ritual Infernet SDK.
Frenrug is a bot in the friend.tech chatroom. Any Frenrug Key holder can send messages to the Frenrug chatroom. In the message, the user can try to convince Frenrug to buy other users' Keys. The Frenrug agent passes the user's message through multiple LLMs running on different Infernet nodes. Initially, all Infernet nodes will be run by Ritual, and then people in the community will participate in running nodes.
Each node responds on-chain to the LLM-generated vote on whether Frenrug should take action. Since LLMs are non-deterministic, each LLM may generate a different response, even if it is the exact same model.
When enough nodes respond, the aggregation request is initiated entirely on-chain. The off-chain Infernet node receives this request and aggregates the various LLM votes into a single action via a supervised classifier and forwards the corresponding validity proof on-chain. The Frenrug proxy contract then performs the action (buy key, sell key, or no action), and the Frenrug key holder sees a reply in the Frenrug chatroom containing the votes of each LLM proxy and the final output.
Strong financing lineup and team background
It is worth noting that Ritual has a strong financing lineup. Robot Ventures, a participating investor, has a high success rate. Among its investments, there are many star projects in every bull market, such as Optimsim, Compound, Lido, Eigenlayer, etc. Hypersphere Ventures is also an investor in Worldcoin and Sei network. In public information, we can also see the participation of famous angel investors such as Balaji Srinivasan, the former chief technology officer of Coinbase.
Ritual’s advisory team is also “star-studded”, including NEAR Protocol co-founder and Transformers co-founder Illia Polosukhin (“Attention is All You Need”), EigenLayer founder and partner Sreeram Kannan, etc. On January 10, BitMEX co-founder Arthur Hayes announced that he had joined Ritual as an advisor.
The Ritual team has many years of experience working in the crypto space. The founders are Niraj Pant and Akilesh Potti. Niraj Pant holds a bachelor’s degree in computer science from the University of Illinois at Urbana-Champaign. Prior to founding Ritual, Niraj was a general partner at Polychain Capital and also served as a board member of CoinDCX, a Sequoia Capital ambassador, and co-founder and CTO of Source Networks. Akilesh Potti graduated from Cornell University and was also a partner at Polychain.
In addition, Anish Agnihotr, a founding member of Ritual and an independent researcher at MEV, also worked at Paradigm as a researcher.
Ritual has not yet issued a coin, but will open source its AI workflow and infrastructure in the coming weeks.
