Partnered With:

Research notes on the Crypto AI project: HyperspaceAI.

Read this article in 9 Minutes

Is decentralized AI a false proposition?

Original title: "ai x crypto research series: hyperspaceAI"

Author: binji

Translated by: 0x26

Editor's note:

In this article, binji, from the Optimism Foundation, explores the potential of combining AI and Crypto using the peer-to-peer AI network HyperspaceAI as an example. The article explains why decentralized AI is necessary, how cryptographic primitives such as fraud proofs and reputation penalties are applied to decentralized AI systems, and how HyperspaceAI will build an Android ecosystem for the AI era.

Why decentralized AI?

The AI revolution is at a crossroads. Today's environment is like a public square that has suddenly turned into a bureaucratic maze of rules and restrictions. Many people are starting to feel that these security measures will become suffocating. OpenAI's AI expert Andrej Karpathy has spoken out about "thinking about centralization and decentralization these days".

BlockBeats Note, Andrej Karpathy previously worked at OpenAI as the head of Tesla AI. In 2017, he left OpenAI at the invitation of Musk to join Tesla as the Director of AI and Autopilot Vision, during which he led the development of Tesla's FSD (Full Self-Driving) system. Elon Musk also replied to Karpathy's post, saying "me too".

Understand Hyperspace

Hyperspace aims to reopen this square and allow everyone to regain this public space. It advocates for the restoration of open collaboration, encouraging developers and AI enthusiasts to return to an environment where their intelligence can freely grow and develop. In the Hyperspace ecosystem, ordinary people and giants have equal power in innovation. We look forward to establishing a broad and dynamic platform for the exchange of ideas, data, and computing power.

The reason why I am interested in this protocol is because the Hyperspace project plans to use encryption primitives such as fraud proof and reputation punishment to ensure trust and disputability among participants. I have always believed that the core primitives of L2 will become more widely used trust primitives on the Internet, not just for proof between L1 and L2. It is exciting to see new ideas emerging that have the potential to further expand the scope of possibilities.

The core operation principle of Hyperspace is similar to a connected delivery system, where each delivery person has a map and a personal seal to ensure the safe and reliable delivery of data or AI tasks, which are referred to as "packages". There is no central node directing the network, instead, these "delivery persons" form a reliable chain to directly transmit information between participants.

Anyone who wants to participate needs to set up a node with computing power and follow the protocol guide to connect to the Hyperspace Community Server (HCS). These servers coordinate the network, match computing tasks with the appropriate nodes, and oversee the verification process.

Hyperspace has the following characteristics that set it apart:

·Decentralized verification: Each node contributes to AI model inference, and the results are cross-verified by other nodes using blockchain-based fraud proof and questioning mechanisms.

·Economic Incentives: Nodes receive cryptocurrency rewards for accurate computation results, and are punished if fraud is detected, ensuring that node interests are aligned with network integrity.

·Open participation: Anyone with computing resources can become a node, which gives everyone an equal opportunity to participate regardless of their personal or company's actual resources. Of course, this is currently only a hypothesis, and if the cost of participation is too high, participation will still be limited. However, compared to the current AI situation, it is still more open.

Blockchain Support: Smart Contract Coordination and Questioning - Responding to Protocol Execution. This means that punishment for non-compliance does not require human involvement, which is important considering the errors that exist within human judicial systems!

Hyperspace uses encryption technology in the following way:

·Unique Node Identity: Nodes obtain a unique NodeAddress through encrypted hashing for secure identification. These addresses are generated using Proof of Work (PoW) and attacks are suppressed by imposing computational costs.

Layered message authentication: Weak signatures are used for non-critical communication, covering elements such as IP addresses and timestamps. Strong signatures are used for critical messages to protect content integrity and prevent attacks.

·Blockchain-based accountability: Immutable ledger records all operations and transactions, providing a transparent stage for truth, and more importantly, everything is disputable! Smart contracts manage fraud proof, enabling on-chain verification of challenges and responses. If suspicious activity is detected, entities can submit on-chain challenges within a specific time window. If inconsistencies arise, violating nodes will be economically penalized, while honest participants will be compensated.

benji Note: I am curious about what specific fault-proofing model they are using.

I am glad to follow their development in this increasingly mature field. Undoubtedly, they will face many difficulties, but the path to accelerating decentralization is not to fight against difficulties, but to master them and brave the wind and waves.