ChatGPT is ONE! How Far Has it Come in Medicine and Healthcare?

Thank you for reading NewHealthcare Platforms' newsletter. We focus on the massive transformative impact of exponential technologies, platform business models, and care delivery innovation on healthcare. In addition, you'll find updates on our validated products and services, designed to facilitate partnerships between startups, tech companies, and healthcare organizations in this rapidly evolving landscape. Don't forget to Subscribe and Follow!

DISCLAIMER: This newsletter contains opinions and speculations and is based solely on public information. It should not be considered business or investment advice. The banner and other images included in this newsletter are AI-generated FAKES created for illustrative purposes only. All brand names, logos, and trademarks are the property of their respective owners. At the time of publication of this newsletter, the author has no business relationships, affiliations, or conflicts of interest with any of the companies mentioned.

Hello again friends and colleagues,

It is hard to believe that ChatGPT just turned one on November 30, 2023. It's a moment to reflect on the extraordinary progress over the last year (and some recent drama), and the profound impact this technology has had on healthcare and medicine. In just one year, ChatGPT has evolved from a promising AI tool to a significant force in medical settings, promising to reshape how healthcare providers practice medicine and interact with patients.

At the forefront of this evolution is the utilization of AI in EHR systems, as exemplified by Epic, the leading electronic health records company. Epic's innovative use of AI and ambient listening technology is revolutionizing providers' workflow as recently shared by Sumit Rana, EVP, R&D.

By enabling physicians to generate progress notes directly from patient-provider conversations and assisting with medical coding, AI is streamlining clinical workflows and enhancing documentation accuracy. Additionally, AI-generated draft responses to patient queries are improving efficiency and fostering more empathetic communication between healthcare providers and patients. Epic’s approach exemplifies how generative AI is being integrated into the healthcare system, providing crucial support to medical professionals and improving patient experiences.

This integration signifies a leap towards a future where AI tools like ChatGPT not only assist in administrative tasks but also contribute meaningfully to patient care. As we celebrate this milestone, let's delve into the latest breakthroughs in AI application within healthcare, focusing on a recent article about GPT-4 and how a new prompting strategy called Medprompt is allowing it to outperform even medically fine-tuned models like Google's MedPaLM 2.

The Cutting-Edge of AI in Medicine: GPT-4 and Medprompt

A recent study titled "Can Generalist Foundation Models Outcompete Special-Purpose Tuning? Case Study in Medicine" sheds light on how a special prompting methodology that the researchers are calling MedPrompting is allowing GPT-4, a generalist foundation model, to outperform specifically fine-tuned "specialist" models such as Google's MedPaLM 2. It is hard to overemphasize the importance of this finding if it validated and confirmed in real-life settings as opposed to the standardized test that were used in the study. It was believed that organizations with proprietary data that is used to fine-tune AI models possessed a competitive advantage since they would possess better performing AI models. But now, it appears that generalist models possess latent performance that can be released by better prompting!

Here is a summary of the findings of the study:

Medprompt: A New Era in Prompt Engineering

The study introduces 'Medprompt', a novel prompting strategy that amalgamates a few advanced prompting techniques (namely, dynamic few-shot example selection, self-generated Chain of Thought [CoT], and ensemble techniques). This innovative approach has shown to significantly enhance GPT-4's problem-solving abilities in medical contexts, outperforming other models on various benchmarks.

Evaluating Medprompt's Effectiveness

Medprompt's performance in biomedical datasets from the MultiMedQA suite was thoroughly evaluated, revealing a remarkable improvement in handling 'eyes-off' data. This suggests that Medprompt can effectively adapt to new, unseen problems without overfitting - a crucial attribute in the dynamic field of medicine.

The graph below from the article, shows how generalist models compare to fine-tuned ones. In the left section, in addition to noticing that GPT-4 with Medprompt is the highest performing model currently, it is astounding to see that a new dot is being added every few months resulting in performance improvement from 40% accuracy in September 2021 to 90.2% accuracy in Sept 2023!! It's been my belief that by early to mid 2024, we will achieve superhuman accuracy on all medical tests. I reiterate that this doesn't mean that AI will replace human providers, but it does mean that not using AI to support medical decision making will likely be considered legally and ethically suspect!

Addressing Limitations and Forward Outlook

While the results are promising, the study acknowledges the inherent limitations and risks, such as the potential for biases in model output. It emphasizes that while GPT-4 excels in benchmark tests, translating these capabilities into real-world medical efficacy requires cautious optimism and further exploration.

The findings from this study are a testament to the evolving capabilities of AI in healthcare. Medprompt, as a tool, unlocks new possibilities for AI application in medical problem-solving, setting the stage for more nuanced and efficient healthcare delivery.

Projecting the Future of Care Delivery and Healthcare Platforms

As we reflect on the transformative potential of GPT-4 and Medprompt, it's imperative to consider how these advancements will shape the future of care delivery and healthcare platforms, as discussed in previous editions of the newsletter some of them are linked at the bottom of this page.

Revolutionizing Patient-Centric Care: With the advent of AI tools like Medprompt, we are moving closer to a healthcare paradigm where personalized, patient-centric care becomes a reality. AI's ability to process vast amounts of patient data and medical literature can lead to more accurate diagnoses and tailored treatment plans, significantly enhancing patient outcomes.

Transforming Healthcare Platforms: The integration of AI into healthcare platforms is not just about technological advancement; it's about redefining the very essence of healthcare delivery. Platforms can evolve to become more adaptive, responsive, and efficient, offering seamless patient experiences. The synergy between human clinicians and AI tools like Medprompt will empower healthcare platforms to offer services that are more aligned with individual patient needs.

Enhancing Decision-Making: The precision and efficiency brought by AI in diagnostic processes will redefine clinical decision-making. Healthcare providers can leverage AI's analytical prowess reducing the likelihood of errors and improving treatment efficacy.

Improving Accessibility and Equity: One of the most significant impacts of AI-driven healthcare platforms will be on accessibility. By democratizing healthcare information and decision-making tools, AI has the potential to bridge gaps in healthcare delivery, making quality care accessible to diverse populations. Multiple research studies have shown that providing AI tools to professionals narrows that performance gap between practitioners by improving the quality of work of low performers to approach that of the highest performers.

Preparing for a Hybrid Healthcare Environment: The future of healthcare will likely be a hybrid model where AI and human expertise coexist and complement each other. Healthcare professionals will need to adapt to this new paradigm, embracing AI as a tool that enhances rather than replaces their skills.

Embracing Continuous Innovation: As demonstrated by the capabilities of Medprompt, the healthcare industry must maintain an environment of continuous innovation and learning. Staying abreast of technological advancements and integrating them into healthcare platforms will be crucial for organizations aiming to lead in the healthcare sector.

To conclude, the intersection of AI and healthcare is not just a point of technological innovation; it's a gateway to a new era of healthcare delivery. As we navigate this exciting journey, it's essential to consider both the immense possibilities and the ethical responsibilities that come with integrating AI into healthcare.

If you enjoyed today's newsletter, please Like, Comment, and Share.

See you next week,

Sam