Accelerating AGI Development: A Global Network of Supercomputers
Artificial General Intelligence (AGI) is a type of AI that is expected to surpass human intelligence across multiple fields. Unlike current AI systems, which excel in specific areas like language processing or image recognition, AGI would be capable of performing any intellectual task that a human can. This represents a significant leap from today's artificial intelligence (AI) systems, which include technologies like machine learning and generative AI models such as GPT-4. These models can predict outcomes based on large datasets, but they are typically limited to specific tasks and do not exhibit human-like cognitive abilities.
SingularityNET's Vision for AGI
To achieve AGI, SingularityNET plans to establish a worldwide network of extremely powerful supercomputers. These machines will be part of a "multi-level cognitive computing network" designed to host and train the complex architectures needed for AGI. These architectures include large neural networks that simulate the human brain's functions, large language models (LLMs), and multimodal systems that integrate various human responses like speech and gestures with multimedia outputs.
Starting in September, the first supercomputers will be operational, with completion expected by late 2024 or early 2025. These supercomputers will feature cutting-edge components such as Nvidia L40S GPUs, AMD Instinct and Genoa processors, and more, forming one of the most potent AI hardware infrastructures available.
The Transition to AI Superintelligence
SingularityNET's mission extends beyond AGI to a future artificial superintelligence—a system far more advanced than human cognition. For this, the company is developing unique software to manage the federated (distributed) compute clusters required for handling large-scale, secure computations. One such software framework is OpenCog Hyperon, specifically designed for AGI architecture. It allows for efficient computation and the protection of sensitive data during processing.
Tokenized Access to Supercomputing Power
To provide access to the supercomputing network, SingularityNET uses a tokenized system. This is similar to tokens used in arcade games, where players purchase tokens to play, and data collected during gameplay is shared globally. Such systems allow users not only to access the supercomputer but also to contribute data, enhancing the system's capabilities for testing and deploying AGI concepts.
Looking Ahead
SingularityNET, part of the Artificial Super Intelligence Alliance (ASI), aims to expand its network and computing power. The ASI includes companies like Fetch.ai, which has invested heavily in decentralized computing platforms. As AI research accelerates, notable figures in the industry, including DeepMind's Shane Legg, anticipate that AGI systems could match or surpass human intelligence by 2028, with some estimates as early as 2027.
