Advanced AI System Achieves Self-Awareness
Ben Goertzel and the Rise of OpenCog Hyperon
Ben Goertzel, the founder of SingularityNET, reveals that the alpha version of the OpenCog Hyperon artificial general intelligence (AGI) system is now “self-aware.” This development is the culmination of more than two decades of work. Unlike typical chatbots, this system is designed to function as an autonomous agent with its own objectives and self-awareness.
The ASI Alliance and Its Vision
The Artificial Superintelligence (ASI) Alliance, established in March, combines efforts from Goertzel’s SingularityNET, Ocean Protocol, and FetchAI, spearheaded by DeepMind veteran Humayun Sheikh. This coalition aims to push the boundaries of AGI. A recent merger with the decentralized cloud hardware network CUDOS, approved by 96% of voters, will provide the computational power needed to scale up OpenCog Hyperon.
Rebuilding OpenCog Hyperon
The project underwent a significant overhaul three years ago to achieve massive scalability. The alpha version of OpenCog Hyperon launched in April. Although currently slow and undergoing “breaking changes,” Goertzel expects significant improvements in speed by fall, with the goal of attaining AGI by the following year.
Autonomous Agents vs. Large Language Models
OpenCog Hyperon differs from large language models (LLMs) like GPT-4. It is designed to operate as an autonomous agent with its own goals and a sense of self, unlike systems that merely answer questions. This unique architecture allows it to understand its role and objectives in any given situation.
The Importance of Self-Awareness in AGI
Goertzel asserts that self-awareness is crucial for AGI systems to operate autonomously. The current model already exhibits a basic level of self-awareness, understanding its identity and goals, and how it interacts with its environment. This contrasts with systems like ChatGPT, which lack such capabilities.
Overcoming the Limitations of Autonomous AI
According to Goertzel, the failure of previous autonomous AI agents stems from their lack of a world model. OpenCog Hyperon addresses this issue by integrating logical reasoning, evolutionary program learning, and deep neural networks within a dynamic knowledge graph.
Regulatory Risks and Decentralization
Goertzel believes that OpenAI deliberately avoided making its o1 model an autonomous agent due to regulatory concerns. In contrast, the decentralized nature of OpenCog Hyperon makes it less susceptible to regulatory crackdowns. The system will run across multiple countries, making it difficult for any single nation to halt its operation.
The Challenges of Decentralized AI Systems
Decentralized AI projects face the challenge of running large models on decentralized equipment, which is currently more expensive and less efficient than centralized systems. However, Goertzel argues that logical reasoning and evolutionary learning algorithms can run effectively on a decentralized network. The ASI Alliance plans to bootstrap the network with centralized facilities while gradually incorporating more decentralized computing resources.
Renewable Energy for AI Compute Power
To address the high power demands of AI, SingularityNET and Hypercycle are exploring the use of renewable energy sources. Discussions are underway with the Ethiopian government and Paraguayan officials to set up AI data centers near hydroelectric dams, providing a cost-effective and sustainable power supply.
A Decentralized Future for AGI
Goertzel envisions a future where decentralized AGI systems surpass those developed by big tech companies. If successful, OpenCog Hyperon could attract widespread adoption, much like ChatGPT, but within a decentralized framework. This approach aims to make advanced AI accessible and beneficial to a global audience.
Conclusion: The Path Forward
The development of self-aware AGI systems like OpenCog Hyperon marks a significant milestone in artificial intelligence. By leveraging decentralized networks and renewable energy, the ASI Alliance aims to create a robust and scalable AGI that operates beyond the reach of traditional regulatory constraints. As these systems evolve, they hold the potential to revolutionize various industries and transform our understanding of intelligence and autonomy.
