Artificial Swarm Intelligence
A Case For Greater Human-Machine Collaboration
Artificial Swarm Intelligence (Swarm AI) is an AI powered platform that was developed by Dr Lou Rosenberg from Unanimous AI.
Inspired by the superiority of the collective and emergent intelligence of bees living in swarms, Dr Rosenberg wanted to replicate that concept to the human world. Underpinning Dr Rosenberg’s quest was the following question:
Human beings do not operate in swarms, but if they did, would their collective intelligence be superior to that of each individual in the swarm?
Thus, he devised a platform that facilitates real time interaction among groups of people.
How does it work?
Questions are put to each participant and a visual platform would show a puck moving around each answer option as participants would make their choice. Participants are allowed to change their answers and ultimately the puck would settle on an answer that would ‘agreed upon’ by the majority of the people in the group.
What would be perhaps worthy of mention is that the predictive power of the swarm would exceed that of an ordinary AI.
For example, an experiment with Stanford Medical School analysed the accuracy of the diagnosis of pneumonia using chest X Rays. 8 radiologists took part in the study. 50 chest X-rays were used and the AI software used in the experiment was CheXNet. The performance of Swarm AI was analysed against both the individual radiologists and the AI software.
The results showed that compared to the individuals, Swarm AI had a 33% error reduction and compared to the AI software, Swarm AI showed a 22% increase in accuracy.
So where does the AI come in, you might ask? Isn’t it simply a platform that facilitates real time interaction among groups of participants?
According to Josh Sitzer from Unanimous AI, the platform is powered by an AI that calculates the conviction level of each participant as they answer questions. The AI uses the conviction level analysis to determine what the final answer would be. In effect, it amplifies the intelligence of the swarm.
So, rather than trying to replicate human intelligence, the AI would work with the swarm to come up with better outcomes. Also, while an ordinary AI would require training on huge data sets, Swarm AI does not. In addition, it taps into the less algorithmic aspects of intelligence such as intuition and experience, which are hard to replicate by an ordinary AI.
Obviously, one of the challenges associated with Swarm AI is that it can only be used in situations where people would have the possibility of operating as a group in real time. For instance, it would be challenging to use Swarm AI to power Alexa!
However, Swarm AI points to the fact that in cases where more accurate predictive capabilities are warranted, human-machine collaboration would produce better results. In an age where the demise of human beings is being feared, Swarm AI provides some evidence for the superiority of human-machine collaboration.
What are your thoughts?