Updated: Apr 26, 2020
Explanations of Blockchain, Artificial Intelligence (AI), & Machine Learning (ML) and how they relate to one another within the context of healthcare
As I continued through my journey of exploring the blockchain landscape while reconciling its fundamental elements against challenges I saw in #healthcare, it didn't take long before I came across articles discussing #ArtificialIntelligence (AI) and #MachineLearning (ML). Stumbling into #AI and #ML added two more abstract and complex concepts to my already steep #technology learning curve. But I couldn't resist the urge to peek into AI and ML to learn more. I would, however, need a simple way to frame blockchain, AI, and ML in my mind for easy reference, and to me a body of water resonates the most.
Blockchain Explained - Under the Surface
The variety and volume of explanations for distributed ledger technology or blockchain is growing rapidly. Yet there are few examples of really simple ways to understand what blockchain is and why it is worth studying. The simplest explanation I have seen comes from Madeline Mann, former Director of People Operations at Gem. She explains blockchain as follows:
"Imagine you are back in first grade. Your teacher hands out gold stars for good deeds. The teacher makes one kid in charge of keeping a tally in her personal notebook of how many gold stars each student has. You all grow uncomfortable with this…is she giving herself more gold stars? She was absent the day I got my gold star, was it recorded? How accurate is her account? You all decide to change the system from centralized to decentralized. Everyone takes out their notebooks and copies down her account of the class’ gold stars. From now on EVERYONE in the class writes down when someone gets a gold star. Now everyone has their own record of truth and no one can cheat and give themselves more gold stars because when their amount is compared to what everyone else has recorded then it will be proven as a faulty star. This is the principle behind blockchain, it is an archive of actions that is spread across thousands of computers so that it is near impossible to tamper with the information."
And wouldn't you know, there are in fact schools beginning to leverage blockchain technology to record, verify, and track students credentials and academic achievements. This concept in academia supports my previously published opinion that credentialing and enrollment functions may be the best place to start applying blockchain technology to make a big impact in healthcare. Credentialing and enrollment are critical first steps for payors and the government authorities to acknowledge clinical providers for compliance and reimbursement purposes. Why couldn't a provider's professional history be logged in an immutable and transparent resource for improving onboarding speed and efficiency?
Hopefully the diagram above, provided by DHL, helps simplify the blockchain process as well. The process of blockchain and the capabilities of AI and ML could create even richer possibilities. Let's continue reviewing these other two computer science topics further to see how.
Artificial Intelligence Explained - SeekingAlpha.com
Artificial Intelligence (AI) explained - above the surface
AI and ML are often used in the same sentence or in place of one another. However, AI is a field of study within computer science and ML is a sub-topic within AI. In other words, ML could be considered an application of AI.
Let's first start with the broad concept of AI. Almost ubiquitous in our day-to-day lives, we have several touch points where we interact with AI inspired applications to make our lives more efficient: Apple's Siri, Amazon's Alexa, Google Assistant, and Microsoft's Cortana are the ones that may reside in your home that help seek out information and assist with some computer-based tasks. My wife and I really appreciate Alexa helping add items to our grocery list application all by voice command.
AI can also inspire applications that help broaden our exposure to different interests. If you've taken up Spotify, Pandora, or Apple Music on their selection recommendations of other songs or artists similar to ones you've played in the past, you've benefited from what the elements of AI can accomplish. These are examples where our voice or our behavior and preferences can be leveraged by technology to make our lives more efficient and broaden our horizons all as a result of the advancing capabilities that AI provides.
If you happen to have an Alexa and have examined the Alexa app you may see a full history of your interaction with the Amazon devices in your home. A view of this history along with some preliminary knowledge of blockchain technology might make the interaction between these two types of technology even more interesting. What may be discussed between you and your smart home devices could be feasibly registered on a blockchain. If you blend smart contracts into the mix this combination might open up some even more interesting possibilities. Say for example you ask Alexa to order an item for purchase and there are commitment expectations within a smart contract for when payments are disbursed based on the performance of the supply chain. Package tracking, that we have with most carriers today, also being a part of the blockchain would help ensure you only pay if the delivery of the product you ordered by voice was within expected timeframes.
Machine Learning Explained - Forbes.com
Machine Learning (ML) explained - above the surface but within the realm of AI
Having just explained AI with a couple household name applications along with a simplified supply chain process, let's now explore ML. The recommendations that may come up from your behavior or preferences goes beyond AI. The mechanisms that allow for computers to study your choices and evaluate these against patterns the machines have been gathering is really ML. Machines, computers with the power and bandwidth to process many thousands of iterations of calculations or algorithms in a relatively short timeframe, can begin to learn from the data and information we have been providing them and increase their understanding of our world.
ML is actually a collection of technological approaches including the following, with some simplified explanations:
Regression is an ML technique used to analyze optimum pricing such as home values based on specific criteria like location alongside bed and bath house configuration.
Classification may require a bit more training during the machine's learning process where the computer develops an understanding of which anomalies may present themselves on your credit card transaction list as fraud. This example may be how your credit card company is applying ML.
Clustering is an ML technique that does not involve training during the machine's learning process. In this scenario the machine is looking for similarities across data. Analyses involving credit risk evaluation might be where this type of ML has been applied.
Anomaly detection in contrast to clustering, is the approach a machine might take to look for differences rather than similarities. For those of us in healthcare, medical billing fraud is a common place where anomaly detection ML is being applied. For example, the machine might review a claim and see that a certain number of medical procedure codes were listed on a claim during a particular amount of time. However, the machine may pick up on the fact that the volume of medical procedures are unrealistic to be performed by a clinician during the time allotted.
Association rules are at play whenever you have seen a "you might also be interested in this" or "others have also purchased these items" as messages that appear on your computer screen just after placing your order on Amazon.com. The computer servers have accumulated the transaction data on a grand scale and begin drawing correlations based on individual's choices.
Time series analyses are a machine's ability to evaluate data across a time period allowing the computer to, as an example, forecast household energy consumption. If you have a Nest smart thermostat, this technique gives the software the ability to begin suggesting certain temperatures at certain times of day or week based on your past adjustments to the dial to get the temperature just right.
Neural networks is perhaps the most abstract technique of all these listed. These networks may be solving the most complex functions of machine learning. Such a network might be trained over time to recognize handwritten numbers. Perhaps the foundational technology behind Optical Character Recognition (OCR) found in healthcare scanning workflows as well as electronic check deposit via image scanning are two examples of this ML application.
With the ML techniques described above along with the potential volume of data on blockchain platforms, computers in any of the nodes on the blockchain network could feasibly process data in much more efficient ways revealing insights we've never seen before. Failing fast, is a description for trying a computation and learning quickly from the failures along the way while using these failures to refine an understanding. Machines have the advantage of failing very fast with their processing capabilities which is an important element of ML. The shared information sequentially logged on the blockchain with computers that have strong AI and ML capabilities could bring to life many advancements that would benefit our personal and business workflows.
Bringing Blockchain, AI, & ML Together
Let's return to the body of water analogy presented at the beginning of this post. What is above the surface are the interactions with computers we can visibly see or audibly hear. What lies below the surface may very well be the processing of the ML, but most certainly the majority would be blockchain technology elements, much less visible. It should be easier to see from the ML examples above that we might experience the AI and ML process because we will actually be providing information to a computer and receiving responses or results back from the computer in exchange. This may reinforce the abstract and hidden nature of blockchain technologies as there is less visible interaction with this platform. Perhaps only the benefits of the attributes that make blockchain appealing and successful are felt by users. The interactions and results of AI and ML are experienced with interactions from Siri, Alexa, and web browsing to name a few. It is feasible that under the figurative surface, blockchain technology could be accumulating the transactional data that the AI and ML assist in generating.
Here is a great video with a helpful diagram where David Houlding of Intel describes the computer science technologies mentioned in this post. He stacks the concepts in a logical fashion and provides helpful explanations of how they are leveraged by one another:
Blockchain in Healthcare, A Layered View - David Houlding, YouTube
If you found this information helpful would you help us out by sharing this post via LinkedIn, Twitter, or your preferred information sharing platform. We would really appreciate your help keeping the blockchain industry dialogue going.
Thank you for reading this post.