A new global network of supercomputers, designed to fast-track the development of artificial general intelligence (AGI), is set to come online in September. Built by SingularityNET, this network aims to push the boundaries of AI beyond current capabilities, with the first node scheduled to be operational within weeks.
AGI represents a hypothetical future where AI surpasses human intelligence across multiple domains, learning from itself and improving its decision-making autonomously. Unlike current AI systems, which excel in specific tasks but fall short in cognitive reasoning, AGI would be capable of comprehensive, cross-disciplinary intelligence.
The supercomputers will form a “multi-level cognitive computing network,” integrating advanced AI technologies such as deep neural networks, large language models (LLMs), and multimodal systems that connect human behaviors with multimedia outputs. This infrastructure will support the development of AGI by providing the necessary computational power and data access.
SingularityNET’s modular supercomputer will feature cutting-edge components, including Nvidia L40S GPUs, AMD Instinct processors, and Tenstorrent Wormhole server racks with Nvidia H200 GPUs. This setup, expected to be completed by late 2024 or early 2025, is designed to handle the massive data and processing demands of AGI development.
Ben Goertzel, CEO of SingularityNET, emphasized the significance of this new infrastructure: “This supercomputer will be a breakthrough in the transition to AGI. We’re witnessing a shift towards continuous learning, seamless generalization, and reflexive AI self-modification.”
The project is part of a broader effort by the Artificial Super Intelligence Alliance (ASI), a collective of companies dedicated to open-source AI research. SingularityNET plans to expand this network in the future, increasing the computing power available for AGI and artificial superintelligence development.
Access to the supercomputer will be managed through a tokenized system, allowing users to contribute data and test AGI concepts. This approach, likened to tokens used in arcade games, facilitates shared access to data sets and computational resources on a global scale.
As the race towards AGI heats up, experts like DeepMind’s Shane Legg and SingularityNET’s Goertzel predict that systems could reach or surpass human intelligence as early as 2027 or 2028. With major players like Mark Zuckerberg also investing heavily in AI infrastructure, the development of AGI may be closer than we think.
