Great article Rubayat. The change to public health is incredible, exciting, scary, and revolutionary. I love the 3 axes you're considering here. Even in the near term, I think the changes you describe are going to compound rapidly. For example, "Thanks to LLMs and other natural language processing AIs, soon we will have ambient listening as a way to capture pertinent data to update patients’ medical history from the raw conversations" might seem like incremental progress to many people. But from 12+ years studying digital health I can't describe what a difference this would make to CHW<>patient interactions. Already far too many CHWs are completely overburdened filling out paper forms, digital forms, and whatever latest sexy-but-underbaked digital tool is in vogue. We once did a time study in Ghana and watched a vaccinator spend 20 minutes delivering a vaccine, of which 18 minutes were spent on manual and digital paperwork. While the hurdles ahead are huge, the rewards could be more so if we navigate these changes well.
Thanks Rubayat for this compiled AI breakthroughs. Generative AI addresses one of the major challenge we have been seeing in AI, especially deep learning algorithms being called black boxes. I spent last year of my PhD in the applications of AI in healthcare just to explain how deep learning models make their decision. I argued a lot that these algorithms can be explained the same way a human doctor does as they mimic the working principles of human brain. Now we are seeing GenAI generating medical reports from their analysis.
Thanks for the article Rubayat -- indeed, gen AI is having a profound impact on healthcare, and I would say, particularly in low-resource settings.
I have been very influenced by Atul Gawande's work on health systems, and in particular, the understanding that in the early 1900s and at all times before that, the limiting factor in healthcare was knowledge about how to diagnose, treat, or manage disease. In 2024, the limiting factor is how to apply -- correctly and efficiently -- the tremendous amount of scientific knowledge we have about how to diagnose and manage illness.
Currently, the biggest barrier to global health equity is that we are unable to consistently apply evidence-based approaches in LMICs the way we are in high-resource settings. Even in well-funded health systems, it's still a huge barrier, but in Kenya where we operate it is a massive problem, which is often summed up as "quality" or "access to quality."
What is really exciting to me about LLMs and their derivatives is that when applied correctly, we can get tremendous improvements in healthcare value without necessarily increasing costs. Through the correct application of national guidelines, we can reduce waste and get way better outcomes even at existing budget levels.
So as much as I am excited in the abstract about your singularity #1, the more urgent equity problem that we need to solve is to level the playing field about getting established treatments to people all over the world who are currently not able to get them.
I really enjoyed reading this, Rubayat. Lots of food for thought about how health technologies, health care, public health, and health systems are likely to change rapidly in the coming years. I appreciate how you have noted both the potential for leap-frogging and also for increased inequities in lower-resource and lower-income settings. For those who are leaders in those settings (or advisors to them), I'd love to hear your thoughts about potential best-case scenarios, likely risks (or negative scenarios), and what we can do in the near future to harness AI for good and get on the best trajectory possible. Looking forward to more conversations about this!!
Thanks Rubayat bhai, for sharing this highly insightful piece that covers a broad spectrum.
To my understanding, the rapid pace and substantial investments in health ai are poised to accelerate specific outcomes, particularly in R&D, training, and distribution. Areas such as ophthalmology, public health, and genetic engineering seem to benefit immensely from these technological advancements. As you mentioned, the ability of ai to pass medical licensing exams and outperform human doctors in diagnostic conversations demonstrates its growing capabilities and potential to revolutionize these fields.
However, the fast-evolving nature of ai technology might bring challenges for many societies. Keeping up with these rapid changes requires substantial adaptability and continuous learning, which may be difficult for some healthcare systems. Additionally, while advanced healthcare systems may seek extraordinary solutions to enhance their long-term efforts, but in many countries the introduction of powerful ai tools in less regulated or non-democratic environments could see an opposite result and increased confusion.
This is particularly concerning given that standards, policies, and protocols are currently set up based on our human abilities and understanding.
Great read, Rubayat. Inevitably, we will see a range of civilization shifts from the singularities you describe and more.
Through collective effort and in due time, we might witness new emergent singularities that seem improbable today (cue nanorobots swimming through our bloodstreams doing continuous telemetry and autonomous course correction in vivo—too wild to imagine today).
These technological shifts remind me of the combustion engine, conceptualized in the 1680s to accelerate human movement. It took a couple of centuries before it was effectively packaged for the masses, with roads, vehicle license regulations, streamlined training, and more.
Today, the velocity of novel technologies in care delivery will only accelerate. But the question remains: when will the packaging and interfaces be ready for the masses? The proverbial road to better care will require a deep interface between the bits and atoms, packaged under the right financial and regulatory sweet spots to effectively pave the way.
I’m really grateful we get to think about this space and potentially make our mark. The singularity may be near or far, but the only thing we can do is take a shot and rest in peace knowing we tried, like our historical pioneers.
A more recent comparison of this movement would be the hype cycle of self-driving cars of the 2010s, now coming to fruition.
Care has come a long way from barbershops being surgery centers.
In conclusion, I look forward to chasing new interfaces to optimize patient outcomes and streamline care processes.
Thanks for an excellent and very informative post, Rubayat. I really enjoyed reading it. The one concern I have relates to the old problem of the deep (evolution programmed) resistance to help-seeking and treatment-adherence amongst most humans. I see a continuing role for AI-assisted human agents far into the future for addressing these concerns and for using the advanced diagnostic and treatment technologies (and protocols) which will need be needed for imaging, sample collection, and treatment of patients. What do you think?
Superb article Rubayat Bhai. It is difficult for even for an AI developer and data processing outfit (read working with several AI heavyweights to fine-tune their models) like Acme AI to keep up with the pace AI is evolving. Particularly medical AI. The only silver lining for moderately technical blokes like us is the bottleneck of clinical trials for these systems that slows the ecosystem just enough for us to hang from the ledge. 😁
On a serious note, we are working with RAG but applied use of LLMs/LAMs tend to still be problematic from a quality response level - and therefore affect usability. Pre-trained models seem to retain a degree of bias and therefore can become problematic without rigorous real-world stress-testing. We mobilised Probahini (a menstrual health and hygiene knowledge assistant working through messenger-based systems) with WaterAid Bangladesh to put a dent in period stigma. Apart from obvious linguistic problems in certain cases, there were a few concerning responses that ranged from 'boring' to 'inaccurate' despite having a vector database. Easily solvable with a bit of prompt engineering and database taxonomy programming though.
Great article Rubayat. The change to public health is incredible, exciting, scary, and revolutionary. I love the 3 axes you're considering here. Even in the near term, I think the changes you describe are going to compound rapidly. For example, "Thanks to LLMs and other natural language processing AIs, soon we will have ambient listening as a way to capture pertinent data to update patients’ medical history from the raw conversations" might seem like incremental progress to many people. But from 12+ years studying digital health I can't describe what a difference this would make to CHW<>patient interactions. Already far too many CHWs are completely overburdened filling out paper forms, digital forms, and whatever latest sexy-but-underbaked digital tool is in vogue. We once did a time study in Ghana and watched a vaccinator spend 20 minutes delivering a vaccine, of which 18 minutes were spent on manual and digital paperwork. While the hurdles ahead are huge, the rewards could be more so if we navigate these changes well.
Thanks Rubayat for this compiled AI breakthroughs. Generative AI addresses one of the major challenge we have been seeing in AI, especially deep learning algorithms being called black boxes. I spent last year of my PhD in the applications of AI in healthcare just to explain how deep learning models make their decision. I argued a lot that these algorithms can be explained the same way a human doctor does as they mimic the working principles of human brain. Now we are seeing GenAI generating medical reports from their analysis.
Thanks for the article Rubayat -- indeed, gen AI is having a profound impact on healthcare, and I would say, particularly in low-resource settings.
I have been very influenced by Atul Gawande's work on health systems, and in particular, the understanding that in the early 1900s and at all times before that, the limiting factor in healthcare was knowledge about how to diagnose, treat, or manage disease. In 2024, the limiting factor is how to apply -- correctly and efficiently -- the tremendous amount of scientific knowledge we have about how to diagnose and manage illness.
Currently, the biggest barrier to global health equity is that we are unable to consistently apply evidence-based approaches in LMICs the way we are in high-resource settings. Even in well-funded health systems, it's still a huge barrier, but in Kenya where we operate it is a massive problem, which is often summed up as "quality" or "access to quality."
What is really exciting to me about LLMs and their derivatives is that when applied correctly, we can get tremendous improvements in healthcare value without necessarily increasing costs. Through the correct application of national guidelines, we can reduce waste and get way better outcomes even at existing budget levels.
So as much as I am excited in the abstract about your singularity #1, the more urgent equity problem that we need to solve is to level the playing field about getting established treatments to people all over the world who are currently not able to get them.
I really enjoyed reading this, Rubayat. Lots of food for thought about how health technologies, health care, public health, and health systems are likely to change rapidly in the coming years. I appreciate how you have noted both the potential for leap-frogging and also for increased inequities in lower-resource and lower-income settings. For those who are leaders in those settings (or advisors to them), I'd love to hear your thoughts about potential best-case scenarios, likely risks (or negative scenarios), and what we can do in the near future to harness AI for good and get on the best trajectory possible. Looking forward to more conversations about this!!
Thanks Rubayat bhai, for sharing this highly insightful piece that covers a broad spectrum.
To my understanding, the rapid pace and substantial investments in health ai are poised to accelerate specific outcomes, particularly in R&D, training, and distribution. Areas such as ophthalmology, public health, and genetic engineering seem to benefit immensely from these technological advancements. As you mentioned, the ability of ai to pass medical licensing exams and outperform human doctors in diagnostic conversations demonstrates its growing capabilities and potential to revolutionize these fields.
However, the fast-evolving nature of ai technology might bring challenges for many societies. Keeping up with these rapid changes requires substantial adaptability and continuous learning, which may be difficult for some healthcare systems. Additionally, while advanced healthcare systems may seek extraordinary solutions to enhance their long-term efforts, but in many countries the introduction of powerful ai tools in less regulated or non-democratic environments could see an opposite result and increased confusion.
This is particularly concerning given that standards, policies, and protocols are currently set up based on our human abilities and understanding.
Great read, Rubayat. Inevitably, we will see a range of civilization shifts from the singularities you describe and more.
Through collective effort and in due time, we might witness new emergent singularities that seem improbable today (cue nanorobots swimming through our bloodstreams doing continuous telemetry and autonomous course correction in vivo—too wild to imagine today).
These technological shifts remind me of the combustion engine, conceptualized in the 1680s to accelerate human movement. It took a couple of centuries before it was effectively packaged for the masses, with roads, vehicle license regulations, streamlined training, and more.
Today, the velocity of novel technologies in care delivery will only accelerate. But the question remains: when will the packaging and interfaces be ready for the masses? The proverbial road to better care will require a deep interface between the bits and atoms, packaged under the right financial and regulatory sweet spots to effectively pave the way.
I’m really grateful we get to think about this space and potentially make our mark. The singularity may be near or far, but the only thing we can do is take a shot and rest in peace knowing we tried, like our historical pioneers.
A more recent comparison of this movement would be the hype cycle of self-driving cars of the 2010s, now coming to fruition.
Care has come a long way from barbershops being surgery centers.
In conclusion, I look forward to chasing new interfaces to optimize patient outcomes and streamline care processes.
Thanks for an excellent and very informative post, Rubayat. I really enjoyed reading it. The one concern I have relates to the old problem of the deep (evolution programmed) resistance to help-seeking and treatment-adherence amongst most humans. I see a continuing role for AI-assisted human agents far into the future for addressing these concerns and for using the advanced diagnostic and treatment technologies (and protocols) which will need be needed for imaging, sample collection, and treatment of patients. What do you think?
Superb article Rubayat Bhai. It is difficult for even for an AI developer and data processing outfit (read working with several AI heavyweights to fine-tune their models) like Acme AI to keep up with the pace AI is evolving. Particularly medical AI. The only silver lining for moderately technical blokes like us is the bottleneck of clinical trials for these systems that slows the ecosystem just enough for us to hang from the ledge. 😁
On a serious note, we are working with RAG but applied use of LLMs/LAMs tend to still be problematic from a quality response level - and therefore affect usability. Pre-trained models seem to retain a degree of bias and therefore can become problematic without rigorous real-world stress-testing. We mobilised Probahini (a menstrual health and hygiene knowledge assistant working through messenger-based systems) with WaterAid Bangladesh to put a dent in period stigma. Apart from obvious linguistic problems in certain cases, there were a few concerning responses that ranged from 'boring' to 'inaccurate' despite having a vector database. Easily solvable with a bit of prompt engineering and database taxonomy programming though.