The concept of superintelligence has gained significant attention, with AI systems on the verge of surpassing human cognitive abilities. This development has raised concerns about the creation of artificial superintelligence (ASI) that may not align with human values. To mitigate this risk, researchers have explored different approaches, such as Constitutional AI and Superalignment. However, there is another promising avenue to consider: Collective Superintelligence (CSi). This approach aims to amplify human intellect by connecting large groups of people into superintelligent systems.
The Power of Swarm Intelligence
Swarm Intelligence, observed in social species like fish, bees, and birds, showcases the capability of collective decision-making without a centralized authority. Inspired by this phenomenon, the idea of connecting human groups into superintelligent systems emerged. Unanimous AI, founded by Louis Rosenberg, pioneered the exploration of this concept by emulating biological swarms using non-verbal techniques.
Through graphical methods, groups collaboratively answered questions and made forecasts, significantly amplifying intelligence. The success of these early experiments led to further validation through academic studies, demonstrating the value of Swarm AI in various applications.
“AI turns $20 into $11,000 on Kentucky Derby Bet.”
– Newsweek
One notable demonstration of Swarm AI’s effectiveness was the accurate prediction of the Kentucky Derby, beating 540-to-1 odds. This achievement showcased the power of connecting a human group into a real-time system that amplified their collective intelligence.
The Evolution of Conversational Swarm Intelligence
While Swarm AI proved valuable for narrowly defined problems, creating a true Collective Superintelligence required a more flexible approach. The breakthrough came with the concept of Conversational Swarm Intelligence (CSI) and advances in AI language models.
Conversational Swarm Intelligence leverages the power of language to enable large groups to deliberate complex issues in real-time. The challenge, however, lies in maintaining thoughtful and coherent conversations as group sizes grow. To overcome this, the concept of hyperswarms was introduced, dividing real-time human groups into overlapping subgroups.
“Can we enable conversational swarms in humans? It turns out, we can by using a concept developed in 2018 called hyperswarms that divides real-time human groups into overlapping subgroups.”
– Louis Rosenberg
AI agents are integrated into these subgroups to facilitate the propagation of information, akin to the lateral line organ in fish. With Conversational Swarm Intelligence, large groups can engage in real-time discussions and converge on solutions amplified by swarm intelligence.
Validation through Experimental Studies
To showcase the intelligence benefits of conversational swarms, researchers conducted a study emulating a famous experiment by Sir Francis Galton. In this experiment, groups were asked to estimate the weight of an ox, and it was found that the statistical mean of the group’s estimates was remarkably accurate.
Similar estimation tasks were carried out using Conversational Swarm Intelligence (CSI), and the results were impressive. CSI outperformed individual human estimations, AI models like ChatGPT, and even the aggregated statistical mean of the group. This significant outcome demonstrated the potential of CSI as a method to achieve superhuman intelligence through collective efforts.
- Survey responses: Average error of 55% (human individuals)
- ChatGPT estimate: Error of 42% (pure AI system)
- Aggregated statistical mean: Error of 25% (Wisdom of Crowds)
- Conversational Swarm Intelligence: Error of 12% (statistically significant)
Potential Applications and Future Possibilities
Conversational Swarm Intelligence technology opens up new possibilities for communication and collaboration on a large scale. From enterprise collaboration to market research and civic engagement, CSI has the potential to enhance various fields. Additionally, as brain-to-brain interfaces continue to develop, the architecture of CSI may enable collective minds to scale infinitely.
Ultimately, while efforts to instill AI models with human values and interests continue to be crucial, alternative methods that amplify human intelligence, such as Collective Superintelligence, should also be explored. With CSI, the goal is not to replace human intellect, but to leverage collective intelligence to solve complex problems while ensuring alignment with human values.
