AI May Improve Healthcare, But Blockchain Will Change It Forever…
Why Distributed Ledger Technology and not AI is the true disruptor in Healthcare
Since IBM announced the firing of 40+ engineers from Watson Health, a number of companies reached out to ask what I thought about it and about AI vs. blockchain solutions in healthcare.
Admittedly, I was already suspicious of Watson Health’s marketing hype and disappointed from it’s failed launches at MD Anderson and others, (mostly due to competition from EPIC) and I was not surprised by its financial under-performance.
However I did not want any ‘Watson prejudice’ to influence my answer to the AI vs. blockchain matchup, so I decided to dive deeper and concluded:
- AI is best thought as Intelligence Augmentation (IA) technology
- AI will augment, not disrupt, the Healthcare business model
- Blockchain will disrupt the Healthcare business model through the network effects of a decentralized economy
#1. AI Is Best Thought As Intelligence Augmentation (IA) Technology
Historically, the phrase “AI” coined in 1956 by John McCarthy, refers to the aspiration of creating a software and hardware entity that possesses human-level intelligence (“human-imitative AI”) and is capable to plan, learn, understand language, recognize objects and sounds and solve problems.
However, most of what is called AI is actually Machine Learning (ML), a subfield of AI, which designs algorithms that processes data, makes predictions and helps make decisions.
(An excellent summary from Vishal Maini can be found here)
Deep learning (DL) is one of the ML approaches inspired by the structure of the brain, called Artificial Neural Networks (ANN). However DL does not ‘think’. DL networks have a hierarchical structure which makes them particularly adapted to ‘learn’ hierarchies of knowledge that are useful in solving real-world problems. For example in image recognition DL analyzes (‘understands’) not just individual pixels, but complex forms like edges, shapes, depth, all the way up to multi-object scenes.
Reinforcement Learning (RL) is another type of ML that does not learn from a training-set of labeled-data, but from interaction with the environment. It does not have direct instructions as to what actions to take and what are the consequences of these actions, but instead ‘learns’ by trying several paths and ‘chooses’ the one that maximizes reward (best action).
(If you are interested in Deep RL, which combines DL and RL, please read this)
#2. AI Will Augment But Not Disrupt The Healthcare Business Model
AI is already used in healthcare. AI-driven apps diagnose patient wounds via smartphone, allow caretakers and doctors to remotely monitor elderly and help digitally verify insurance information.
AI is also predicted to save up to $400 Billion in healthcare through automation and predictive analytics, displacing physical and manual labor with jobs requiring high technological skills. Changing the labor market will require a redesign of the clinical workflow while developing human-machine collaborations (below).
Interestingly, AI will not only ‘learn’ human biases but also amplify them (e.g. ‘assume’ doctors are male and nurses are female). It is imperative to make sure that AI solutions will be ‘trained’ to produce trustful, safe and compliant solutions (Explainable AI) and ‘behave’ responsibly (Responsible AI).
#3. Blockchain, Not AI, Will Disrupt The Healthcare Business Model
(For an excellent article about AI/ML and blockchain please read Jorden Woods post here).
Despite the continual improvement of AI, ML, RL, DRL, creating convolutional neural networks (CNNs) for image identification, recurrent neural networks (RNN) for natural language processing (NLP) and generative adversarial networks (GANs) for mimicking human creative capabilities, there are problems blockchain solves better than AI like securing user privacy, distributing data, protecting datasets from bias, manipulation and hacking and guaranteeing data transparency (think TrueBit).
The problem in essence with AI is that it remains a centralized solution where “data is centrally stored, owned and controlled by the group that collects it”. (Jorden Woods, 2018)
Therefore AI does not and can not fundamentally change the current business model in healthcare because it lacks the ability to create a decentralized, participatory economy of shared knowledge and insights (think Numerai and Ocean).
A fully decentralized and tokenized data exchange system with a transparent reward model will not only be attack-, collusion-, and censor-resistant, but will self-reinforce network effects from data users, providers and scientists. The better the system performs, the more capital it attracts, which means more potential payouts, which attracts more data providers and scientists, who make the system smarter (something that did not happen with Watson).
Final thoughts:
The promise of AI is to help physicians keep their patients healthier, prevent lengthy hospitalizations or re-hospitalizations and identify gaps in care while adhering to best practices.
However, Healthcare does not have an optimization problem where AI simply assists with the management of an information-intensive environment. Healthcare is an opaque, multi-stakeholder, friction-full, misaligned ecosystem that fundamentally lacks trust. Therefore it needs the network effects of decentralization to empower self-sovereignty and transform patients from healthcare consumers into health producers.
“Blockchain is more than just ICT* innovation. It facilitates new types of economic organization and governance... based in new institutional and public choice”… (Economics of Blockchain. De Filippi, Davidson, Potts, 2016)
- ICT- information communication technology
Acknowledgements: Thanks to Jorden Woods and Radhika Iyengar-Emens at DoubleNova Group, for their review and contribution.
If you liked what you read, go ahead and “Clap” below so others will see it too (up to 50 claps allowed!)
Join my upcoming conference call on Healthcare and Blockchain HERE
