This article originally ran on Forbes.com, and is authored by Exponential AI’s Chief Revenue Officer, Florian Quarré.
Examine just about any technology news source in 2018, and you will likely find a number of business leaders considering what artificial intelligence and blockchain technologies might do for their sector. It makes sense: Blockchain, with its roots in cryptocurrency, has emerged as a disruptive and innovative new way to enable secure and efficient transactions for many industries. Moreover, the technologies that power artificial intelligence are already driving insights across a variety of sectors.
In my field of helping health care organizations utilize emerging technologies, I believe AI and blockchain could converge to create something exponential. At the same time, as we in health care consider the technical and ethical implications of health care records stored in some capacity on the blockchain, we are already making extensive use of cognitive technologies, allowing the systematic execution of ambiguous tasks that once required human effort and interpretation. Cognitive technologies — the domains of AI — that have seen rapid progress and investments include computer vision, natural language processing, machine learning and speech recognition.
Although no AI application has yet reached a state of artificial general intelligence, there are numerous successful examples of applied AI that, once brought together, deliver high accuracy in information treatment and workforce augmentation.
In my experience as a chief digital officer, one such example of applied AI considers the use of cognitive technologies to acquire, structure and deliver health records into research and enterprise-grade data sets, which has helped break down the illiquidity that has made data challenging to share in the health care industry. Computer vision is used in coordination with natural language processing to convert the millions of records and notes that transit in the form of images into electronic documents; these electronic records are then worked on with a variety of ontologies to disambiguate the information they contain to reach a high level of accuracy and eventually reach a point of structured digital record. The workforce reviewing the digitization of these records is guided through the review rather than required to manually recapture the data contained. It makes adjustments to the information that was incorrectly transcribed by the system, which systematically feeds a machine learning feedback loop to continuously train the system with manual intervention for greater accuracy.
However, I believe applying AI to extract and structure information held in electronic medical records is only a first step. Where AI offers value in the near future is in interaction with a health care blockchain, where the benefits of distributed records can be combined in sophisticated and powerful ways with cognitive technologies to target the democratization of health insights.
Imagine, for instance, how AI could be deployed on a unified health information blockchain to accelerate the discovery, development and delivery of personalized medicine. If the blockchain were populated with codified, abstracted and distributed health data, with a non-alterable control over privacy, AI could then be allowed by patients and organizations to search those records as a representation of the patient’s longitudinal databank in search of health-related warning signs and red flags, trends and insights, outbreaks and overlooked cures.
While health information accessible by AI via blockchain promises to accelerate a generation of health care insights, there are also worthwhile counter-narratives to consider regarding our privacy of data.
Health information is personal. The health data we generate reflects who we are at our genetic levels, the care we are receiving and the outcomes our care is providing. Although it might be useful for corporations and universities to have that information available into its most distributed format on the blockchain, the transparency blockchain offers, along with the sheer size of medical records, might make it an impossible solution. Yet, with the right model, I believe there is no reason to think that we cannot utilize the broader medical learning potential in the data we are generating in health care, so long as we are not storing the records on the blockchain directly, but rather using blockchain as a mechanism by which we offer access and authorization for medical records.
In the situation of an authorized AI-based agent, for example, an individual could grant permission to an AI agent and inscribe on the blockchain that consent is given to consume the patient’s longitudinal record for early signs of cancer. While that record might not live on the blockchain itself, the permission to retrieve and interpret the record can be achieved with total automation thanks to the adoption of blockchain coupled with the advancements in AI.
Extended further, the same individual could deploy not one, but an army of specialized AI agents, each built and certified by a medical community comprised of leading doctors, researchers and other leaders of the field. The individual could have a high degree of confidence that these automated tools would help find early stage signs and symptoms of diseases by analyzing and contextualizing health data from every visit to any doctor for the rest of their natural life.
I believe AI models are hungry for semi-structured data sets; they can uncover patterns hidden to the naked eye and create meaning in the linkage of thousands of disparate entities. A permissible distribution of health information via blockchain could make available for the first time in history a wide array of executable data sets for these highly specialized, trusted, narrow AI agents, with the ultimate goal not just of aggregating and analyzing data, but of bettering the care delivered to patients everywhere.