HyperCycle, a blockchain project led by CEO Toufi Saliba, aims to decentralize artificial general intelligence (AGI) by connecting AI machines. With over 347,000 computational nodes, it focuses on reducing AI compute costs and enabling seamless transactions.
HyperLoop, a groundbreaking early-phase blockchain initiative sans ledgers, aiming to connect artificial intelligence (AI) systems for resource sharing, has successfully onboarded 347,000 computational nodes onto its network, as confirmed by CEO Toufi Saliba.
Currently, the cost per node stands at a modest $1,000. While the monetary value isn't staggering, the project's vision extends to becoming the "internet of AI," with aspirations to eventually decentralize artificial general intelligence (AGI).
Ben Goertzel, the founder of SingularityNET, serves as the project's chief AI scientist. He initiated this venture due to the inadequacies of existing blockchains in providing the requisite speed and coordination for AI operations. HyperLoop leverages the Toda communication protocol, proof-of-reputation, modified proof-of-work, along with Cardano's Hydra.
"Every transaction achieves finality in just about 300 milliseconds," Saliba revealed to Cointelegraph during an interview held at the Beneficial AGI Summit in Panama this week.
Delving into the conference's theme, Saliba cautioned about the rapid evolution of AGI into a superintelligence, potentially endowing significant power to the first centralized entity or authoritarian regime to develop it.
"It could potentially spell disaster for humanity," remarked Saliba, drawing from his extensive decade-long experience in AI. He firmly advocates for decentralized governance as the sole safe approach towards realizing a beneficial AGI.
Hence, the grandiose ambition encapsulated in the HyperLoop project's name is justified. Presently, it's a business-to-business endeavor primarily focused on integrating AI enterprises. Each HyperLoop node comprises a virtual machine, transaction machine, and AI machine, facilitating an average of 150 AI agents per node to engage in interactions and transactions.
Saliba illustrated the process using an example involving an optical character recognition (OCR) AI encountering text in the ancient Aramaic language. It could seamlessly access an AI service specialized in Aramaic OCR, execute payment via a smart contract microtransaction, and utilize the service instantly, seamlessly integrating it within its own application.
He disclosed that a prominent deepfake video application is already leveraging the network. In a recent podcast appearance, he mentioned a meeting with Nvidia to discuss the concept of the "internet of AI," refraining from divulging further details.
Saliba clarified that the companies deploying these nodes are primarily driven by profit motives or the desire to mitigate their substantial AI compute expenses, rather than concerns about cryptocurrency or humanity's welfare.
"Decentralization isn't a priority for these entities. Most are established AI firms solely interested in cost reduction, aiming to trim down their $200,000 bills to $120,000,"
The scalability of the network remains an open question, given its relatively nascent stage. However, one intriguing albeit theoretical possibility is that interconnecting AI systems to collaborate might pave the way for AGI development. Saliba posits that the network itself could serve as a conduit towards achieving AGI as an emergent property, where the collective intelligence of interconnected AIs surpasses the sum of individual components.
Using the analogy of a colossal brain, wherein each AI node represents a neuron capable of synchronized firing, Saliba expounded,
"The proliferation of such nodes globally increases the likelihood of AGI emergence, presenting the entire world with a collective opportunity to attain AGI."