“Swarm intelligence is the ability to group many minds together to work more efficiently, effectively, and cooperatively in order to achieve progress beyond the sum of its parts. Technology is the framework necessary to group minds together to allow for swarm intelligence to occur in humans. This will lead to a new level of intelligence in humans.” 

Peter Marino, science and technology enthusiast


Present day machine and deep learning face several daunting challenges in clinical medicine: uncertainty, ambiguities, heterogeneity, and if these are not enough to thwart AI technologies in this domain, data privacy legislations.

At present, most health and medicine projects rely on local learning with disconnected locations or central learning with cloud-based machine learning, and a few are considering federated learning with data at the edge but parameter settings orchestrated by a central parameter server.

The authors present the novel concept of swarm learning, a decentralized machine learning approach that cleverly bundles edge computing, blockchain-based peer-to-peer networking and coordination while maintaining confidentiality without the need of a leader or central coordinating body. This swarm intelligence or theory is often manifested in nature with birds in flight or ants in colonies or in humans with participants in a market economy.

The study involved four use cases of diseases with heterogenous presentations: COVID-19, tuberculosis, leukemia, and lung pathologies. A total of 16,400 blood transcriptomes as well as 95,000 chest X-ray images were used with non-uniform distributions and substantial study biases. The authors showed that the swarm learning classifiers outperformed the local classifiers. In addition, the authors were able to demonstrate that the local confidentiality regulations were fulfilled.

The COVID-19 global pandemic has accelerated the need for a reliable, real-time, and secure AI solution that will preserve privacy. This paper illustrates an innovative decentralized approach to the present day centralization of data strategies with both data and parameters at the edge in the form of swarm edge nodes that exchange parameters for learning via blockchain technology.

The myriad of advantages that swarm learning offers can neutralize the set of obstacles in biomedicine, especially that of data privacy and security, so that AI can finally be leveraged to its fullest potential.

A word of caution: even swarm intelligence and learning may not be correct so human intelligence is still necessary after all.


The full article can be read here