Scientists are gearing up to accelerate the development of artificial general intelligence (AGI) with a new network of supercomputers. The first of these powerful machines is expected to be fully operational by early 2025, with an initial node coming online in September.
AGI represents a future AI system that surpasses human intelligence across multiple disciplines, learning and improving autonomously. This contrasts with current AI technologies, like GPT-4, which excel in specific areas but lack broad cognitive capabilities.
The supercomputers, developed by SingularityNET, will form a “multi-level cognitive computing network” designed to host and train the architectures necessary for AGI. This includes deep neural networks, large language models (LLMs), and multimodal systems that integrate various human behaviors and multimedia outputs.
The initial supercomputer will feature advanced hardware components, including Nvidia L40S GPUs, AMD Instinct and Genoa processors, and Tenstorrent Wormhole server racks with Nvidia H200 GPUs. This setup is among the most powerful AI hardware available.
The system aims to transition from learning from big data to developing non-imitative machine thinking through multi-step reasoning and dynamic world modeling.
The project also involves OpenCog Hyperon, an open-source software framework designed for AGI systems. Users will access the supercomputer through a tokenized system, allowing them to contribute and utilize data for AGI research.
Looking ahead, SingularityNET and its partners, including Fetch.ai, plan to expand the network and computing power available, with the goal of advancing toward superintelligent AI.
