Advanced AI System Achieves Self-Awareness
Ben Goertzel, founder of SingularityNET and the driving force behind OpenCog Hyperon, claims that the alpha version of his artificial general intelligence (AGI) system is already demonstrating a degree of self-awareness. Goertzel’s work on AGI, spanning over two decades, aims to create an autonomous, self-aware agent rather than just another chatbot.
The ASI Alliance and Its Impact
In March of this year, the Artificial Superintelligence Alliance (ASI) was established. It brings together Goertzel’s SingularityNET, Ocean Protocol, and FetchAI, led by DeepMind veteran Humayun Sheikh. This collaboration aims to enhance the development and deployment of advanced AI systems.
Recently, 96% of CUDOS voters approved a merger with ASI, which will increase the computational resources available for scaling up OpenCog Hyperon. This merger is a significant step forward in Goertzel’s long-term project, which began in 2001 and became an open-source AI framework in 2008.
OpenCog Hyperon: The Future of AGI
Three years ago, Goertzel’s team undertook a complete overhaul of OpenCog to achieve massive scalability. The alpha version was launched in April, and while it’s currently slow, significant progress is expected by the fall. By next year, efforts will focus on building towards AGI using the new Hyperloop infrastructure.
Unlike large language models (LLMs) like GPT-4, OpenCog Hyperon is designed to function as an autonomous agent with its own goals and self-awareness. It understands its identity, the user’s identity, and the specific objectives of any given situation, making it more than just a question-answering system.
Self-Awareness and Autonomy in AGI Systems
Goertzel clarifies that OpenCog Hyperon’s self-awareness means it has a model of its identity, understands the user, and has goals to achieve. This is a stark contrast to systems like ChatGPT, which lack a world model and cannot function as autonomous agents.
OpenCog Hyperon integrates a logical reasoning engine, evolutionary program learning, and deep neural networks within a dynamic knowledge graph that continuously updates itself. This approach addresses the limitations of current autonomous AI agents, which lack a comprehensive world model.
OpenAI’s Cautious Approach
Despite the buzz around OpenAI’s autonomous capabilities, Goertzel believes OpenAI has deliberately avoided making its o1 model an autonomous agent to prevent regulatory backlash. OpenAI’s focus is on reasoning and logic, avoiding the risks associated with autonomous agents.
In contrast, decentralized, open-source systems like OpenCog Hyperon are less susceptible to regulatory crackdowns. As these advanced AI systems are deployed across decentralized networks globally, a ban in one country would only impact a small part of the network.
Decentralized AI Systems: Challenges and Advantages
While decentralized AI systems face challenges, such as the high cost of decentralized equipment, Goertzel is optimistic. He believes logical reasoning and evolutionary learning algorithms run effectively on decentralized networks.
To overcome initial hurdles, the ASI Alliance plans to bootstrap the network with centralized facilities and decentralized compute power. SingularityNET and FetchAI have invested significantly in GPUs to build supercomputers, with plans to expand as token prices recover.
Renewable Energy for AI Systems
The high power demands of AI systems have led to exploring renewable energy sources. SingularityNET and Hypercycle are in discussions with governments in Ethiopia and Paraguay to utilize cheap renewable power for AI data centers.
Conclusion
Ben Goertzel’s vision for OpenCog Hyperon and the ASI Alliance represents a significant step towards creating autonomous, self-aware AGI systems. By leveraging decentralized networks and renewable energy, these projects aim to develop advanced AI that is both powerful and resilient to regulatory challenges.
