It’s likely that almost none of us had heard the term “Generative AI” a year ago, although it’s a concept that’s been around for about five years. But with the public release of OpenAI’s ChatGPT3 last November, it seems to be all we’re hearing about. ChatGPT writes astonishingly convincing prose and poetry and its cousin DALL-E 2 creates images of almost any subject, in any style (like this cat, portrayed in the style of Picasso – see cover image for this post!). These new technologies are exciting, but they also eerily seem to have a mind of their own (Ghost in the Machine, anyone?).
Intrigued, I thought I would give ChatGPT a try, and asked it the following question:
Why should employers select Crossover Health for care delivery and health cost management?
The following response was generated in seven seconds:
There are several reasons why employers should select Crossover Health for care delivery and health cost management:
- Comprehensive services: Crossover Health offers a wide range of care delivery and health cost management services, including primary care, urgent care, chronic care management, health coaching, and population health management. This allows employers to have a one-stop-shop for all their health needs.
- Proven results: Crossover Health has a track record of success in improving the health of employees and reducing healthcare costs. In a recent study, Crossover Health was shown to reduce hospital admissions by 38% and emergency department visits by 24%, while also lowering overall healthcare costs by 10%.
- Convenient access: Crossover Health provides on-site health clinics at employer locations, making it easy for employees to access care without having to take time off work. In addition, Crossover Health offers virtual care options for employees who prefer to receive care remotely.
- Personalized care: Crossover Health takes a patient-centered approach to care, providing personalized treatment plans and health coaching to help employees achieve their health goals.
- Partnerships with leading health plans: Crossover Health partners with top health plans, such as Aetna, to provide high-quality care for its plan sponsors and their benefitted members. This ensures that employees have access to the most comprehensive primary health delivered with a concierge experience.
This completely computer-generated response is frankly incredible for a first pass. And, my guess is that this technology will only get better with more time (i.e., more context), more training (i.e., even larger datasets), and more technological developments (more generative computing power). Which raises the question of the potential impact of Generative AI in healthcare? What are the possibilities? What are the risks? Which direction does this bend the arc of creativity and ultimately humanity?
Massive datasets have been used for traditional AI-powered health analytics and insights for years, at both a population and individual level. For example, they help insurers more accurately assess the risks in a given population, and allow providers to quickly determine an evidence-based course of action for an individual’s health journey, often based on millions of similar diagnoses and outcomes. Unlike traditional AI, though, the essence of generative AI is that it can create new content in a semi- or non-supervised manner, not just sort and classify existing datasets and patterns. In an interesting use case, Generative AI is being tested as a way to invent new proteins or drugs. It could also mine the millions of provider notes in health records, map these against a patient’s own self-diagnosis, and then generate its own conclusions and recommendations. But I don’t think it will, or should.
We have long talked about the potential of the instrumented patient as a medically supervised version of the quantified self, but much less so about the augmented physician (or care team) with even more intelligence available from ever more data. The applicability of the HITL concept to diagnoses or treatment plans is clear and supports my notion that the doctor won’t be replaced, but instead augmented. But what will be critical is the way that the augmentation takes place.
Despite all the hoopla, what is forgotten is that generative AI is not good at the most important aspect at the core of care—empathy. The concern with the notion of Medicus Ex Machina is best portrayed by the unfeeling Hal in 2001: A Space Odyssey. Despite near infinite intelligence, Hal is never human enough. The human touch is not just tactile, it’s also a tangible concept, the importance of human connection and human relationships cannot be lost amidst the AI hype. The physician’s empathetic concern, understanding touch, and human relatability (plus accompanying fallibilities) will remain critical to the healing process, as well as to the future health enterprise. The cold precision and empty soul of technology will undoubtedly struggle to match the healing power of the human spirit.
While we are early in the Gartner Hype Cycle for generative AI, we should prepare for what’s coming. Everyone is breathlessly focused on what this new technology may replace, including in healthcare. But when the hype dies down, the technology matures, and experience brings wisdom, we’ll be left looking into how it can augment and amplify the care our providers can deliver now; as well as what will be possible in the future. Instead of the steely precision of Medicus Ex Machina, we should be looking forward to the warmth of Hominus Auctus (see next post!)
And that will create the brave new world of health as it should be.
# # #
Post Script: Of interest I have been following some recent articles where AI has replaced humans, seemingly without anyone noticing. The capabilities of generative AI (and ChatGPT in particular) have set alarm bells ringing in education, as increasing numbers of suspiciously well-crafted papers have been crossing teachers’ and professors’ desks. Better yet, how about taking your medical or bar exams for you? While many in education have expressed concerns about generative AI applications, other academics are seeing that tools like ChatGPT (and similar apps, such as Google’s LaMDA and the startup character.ai) are here to stay, and can be a useful resource to raise students’ analytical and creative skills. This is the “Human in the Loop” (HITL) concept, where generative AI is used to develop basic concepts (in software coding, music, ad copywriting, architecture, medicine, etc.) to augment or jumpstart a human’s own creative or professional processes. However, at some point, the wonders these technology advancements bring and the mystery of how they are accomplished unfortunately also evoke the dystopian classics from The Matrix to 1984 to Blade Runner to Snowcrash. But, rather than yielding to that dark future, we have a chance to proactively pull forward the light and goodness of augmented and amplified humanity.
Scott Shreeve, MD 