The Path to Artificial Superintelligence
MIT Technology Review explores the path to artificial superintelligence through the development of a semantic layer that enables agents across domains to work together and 'think' together through shared intent, context, and reasoning.
Intelligence analysis by Llama
The article discusses the need for a semantic layer to enable agents to work collectively, citing the example of a healthcare system made up of multiple AI agents. This layer, called the 'Internet of Cognition,' enables agents to share intent, context, and reasoning, and is seen as a key step towards distributed artificial superintelligence.
Imagine a team of AI agents working together to help people. Each agent is an expert in its own area, but they can't work together yet. A new layer of technology, called the 'Internet of Cognition,' is being developed to help them share ideas and work together more effectively. This could lead to breakthroughs in areas like healthcare and education.
Analysis
The Intelligence is Already There
The article highlights that the intelligence is already present in individual AI agents, but what is missing is the connective tissue that turns them into a cohesive team. Vijoy Pandey, senior vice president and general manager of Outshift by Cisco, explains that this connective tissue comes from adding a semantic layer that enables agents across domains to work together and 'think' together through shared intent, context, and reasoning.
The Internet of Cognition
This semantic layer, called the 'Internet of Cognition,' relies on a connectivity layer beneath it called the 'Internet of Agents,' which allows autonomous agents to discover one another, prove identity, and exchange messages across domains. When used together, they enable 'the next step on the road to distributed artificial superintelligence,' says Pandey.
The Next Axis of Scale
Pandey emphasizes that the next axis of scale must be horizontal, enabling multi-agent systems to work together and share knowledge and reasoning. This is in contrast to the current focus on vertical scaling, which has led to bigger models, trained on more data with more compute. While vertical scaling can produce more capable agents, it does not enable agentic problem solving across different systems, companies, and platforms.
The Gap is Architectural
Pandey notes that the gap between current multi-agent systems and the desired outcome is architectural, not a prompting problem. Without the right coordination layer, naive multi-agent setups can perform worse than a single agent. The step change is that team of agents converging on its own, on a new problem, with no human stitching the seams.
Key points
- The intelligence is already present in individual AI agents, but what is missing is the connective tissue that turns them into a cohesive team.
- The 'Internet of Cognition' is a semantic layer that enables agents across domains to work together and 'think' together through shared intent, context, and reasoning.
- The next axis of scale must be horizontal, enabling multi-agent systems to work together and share knowledge and reasoning.
- The gap between current multi-agent systems and the desired outcome is architectural, not a prompting problem.
The development of the 'Internet of Cognition' could lead to significant breakthroughs in various industries, including healthcare and education. The ability of agents to work collectively and share knowledge and reasoning could lead to more effective solutions to complex problems.
The development of artificial superintelligence also raises concerns about the potential risks and unintended consequences. For example, the ability of agents to work collectively and share knowledge and reasoning could also lead to the creation of more sophisticated and potentially malicious AI systems.
