Healthcare’s Knowledge Problem

Healthcare is becoming a real test of how we sustain knowledge. The challenge is no longer just storing information, but keeping it usable, accurate, and current, while also cutting down the time clinicians spend reviewing charts and writing notes. 

One of the clearest near-term benefits of AI is clinical summarization. Instead of digging through scattered notes, lab results, medication lists, imaging, and visit transcripts, clinicians can get a clear, unified picture of the patient. This is where tools from OpenAI and Anthropic are heading. OpenAI is positioning its healthcare offerings around patient-facing summaries and enterprise systems designed to meet privacy and compliance needs. Anthropic is developing similar healthcare-focused tools and infrastructure, especially for clinical and life-science workflows.

But research shows there is a catch. A recent study on AI-assisted report writing for chronic disease care found that the AI produced high-quality drafts with very few edits and no safety problems. Even so, clinicians spent about the same amount of time reviewing these drafts as they did writing reports by hand. The reason is professional responsibility. In medicine, clinicians feel obligated to check everything carefully, even when the AI is usually right. This creates what researchers describe as an accountability paradox: accuracy alone does not reduce workload if full verification is still required.

Because of this, the real challenge is shifting from asking whether AI can write well to asking how systems can support selective verification. The goal is to let clinicians quickly see what matters, what changed, and what evidence supports each statement, without forcing them to recheck everything from scratch.

Another important development is the push toward better medical memory. Patient information is often scattered across systems, making it hard to trust summaries or recommendations. Efforts to unify labs, medications, visit notes, and recordings into a single, traceable context aim to reduce this fragmentation. When data is well connected and clearly sourced, AI can organize and summarize it without guessing.

Open models are also entering the picture. Google’s medical models, including MedGemma and MedASR from Google, are notable because they support an open, developer-friendly ecosystem. This approach appeals to organizations that want strong medical AI capabilities while keeping local control over data and governance.

Taken together, the pattern is becoming clear. AI that simply drafts text is helpful, but AI that drafts and clearly shows where every claim comes from is far more sustainable. The most promising systems ground their outputs in linked evidence, make data sources and versions easy to audit, and reduce repeated work by improving search, organization, and de-duplication. In healthcare, the most sustainable knowledge is the knowledge clinicians do not have to recreate again and again.

REFERENCES

Lee C, Vogt KA, Kumar S. Prospects for AI clinical summarization to reduce the burden of patient chart review. Front Digit Health. 2024 Nov 7;6:1475092. doi: 10.3389/fdgth.2024.1475092. PMID: 39575412; PMCID: PMC11578995.

Zhang X, Yu J, Yan P, Jiang L, Shen X, Cheng M, Liu X. Human-in-the-Loop Interactive Report Generation for Chronic Disease Adherence. arXiv preprint arXiv:2601.06364. 2026 Jan 10.

https://openai.com/index/introducing-chatgpt-health/ "Introducing ChatGPT Health"

https://www.anthropic.com/news/healthcare-life-sciences "Advancing Claude in healthcare and the life sciences"

https://www.axios.com/2026/01/12/openai-acquires-health-tech-company-torch "OpenAI acquires health tech company Torch"

https://developers.google.com/health-ai-developer-foundations/medgemma/model-card? MedGemma 1.5 model card | Health AI Developer ..." 

[TIME](https://time.com/7344997/chatgpt-health-medical-records-privacy-open-ai/)

[Axios](https://www.axios.com/2026/01/12/openai-acquires-health-tech-company-torch)

[Business Insider](https://www.businessinsider.com/anthropic-chases-openai-ai-heath-claude-2026-1)

[The Economic Times](https://m.economictimes.com/tech/artificial-intelligence/openai-acquires-healthcare-startup-torch-deal-pegged-at-100-million/articleshow/126495784.cms)

Is the Future of Medicine Just a Prompt Away?

In Evaluating General-Purpose LLMs for Patient-Facing Use: Dermatology-Centered Systematic Review and Meta-Analysis (medRxiv, 2025), the data tells a fascinating story: large language models (LLMs) are improving in medical reasoning, empathy, and safety - but they’re not perfect, and trust takes time to earn. Which, come to think of it, sounds a lot like the long human history of hoping for miracle healers.

Long before stethoscopes, scalpels, and sterile gloves, our first “doctors” were magicians - or at least, that’s what everyone believed. Prehistoric healers waved bones, mumbled incantations, and applied sometimes questionable herbal pastes. Yet enough patients recovered to keep the legend alive.

Fast forward a few millennia and not much has changed… except the props. The bone rattle has been replaced by a diagnostic app. The “spirit-cleansing smoke” is now an MRI scan. And our new shamans? They’re called AI engineers.

Just like in the old days, we still crave the miracle cure, the instant fix, the all-knowing healer. Our dream is a tireless personal doctor who remembers every ache, every allergy, every bit of medical literature (plus the plot of every episode of Grey’s Anatomy).

When ChatGPT burst into the public spotlight in late 2022, some were fascinated and some were wary. Could a chatbot really diagnose a rash? Suggest a safe treatment? Explain it all in plain language?

Early studies, including those reviewed in the paper, painted a mixed picture. In 2022, the mood was skeptical. By 2023, optimism surged as newer models like GPT-4, Claude, and Gemini started showing measurable gains in accuracy, empathy, and communication. But by 2025, the mood had shifted again - not to cynicism, but toward a more critical view.

The truth is, AI in medicine is a lot like the magic of old: it works impressively well in certain contexts, but not always when or how you expect. LLMs are now better at interpreting images, offering solid medication safety advice, and even admitting when they don’t know - a kind of digital humility our ancestors probably wished their witch doctors had. But they still have limits. Even when an AI aces a medical board exam and offers great second opinions, patients using it alone don’t necessarily make better decisions.

That’s why the paper calls for evaluator-aware, patient-in-the-loop frameworks - ways of measuring not just whether the AI gets the right answer, but whether it helps real people make better choices. Because in healthcare, as in magic, the spell only works if it actually helps the patient in the real world.

REFERENCE

Irene S. Gabashvili Evaluating General-Purpose LLMs for Patient-Facing Use: Dermatology-Centered Systematic Review and Meta-Analysis medRxiv 2025.08.11.25333149; doi: https://doi.org/10.1101/2025.08.11.25333149

The Past, Present, and Future of AI-Powered Medicine

In this era of rapidly advancing technology, Artificial Intelligence (AI) is spearheading transformative changes, particularly within the healthcare sector. A prime example is ChatGPT, a game-changer that is fast becoming an influential player in the biomedical field. Its potential to catalyze innovation and revolutionize medical research is profound. 

A MedrXiv paper posted today is a systematic review of AI in biomedical literature. The scope of this review is broad, encompassing preprints, peer-reviewed articles, case reports, patents, clinical trials, and even FDA approvals. The paper talks about impact of AI, including ChatGPT, across all medical specialties and subsets of publications, finding overrepresented domains within each subset, highlighting potential research gaps, biases, or areas of excessive focus. 

The pace at which the biomedical literature is growing is staggering, with two new papers being added every minute, around the clock. The MedrXiv paper also brings to light the limitations of current medical publishing models. 

The human author collaborated with ChatGPT, Bing, Claude, and Bard, and used SciSpace Copilot to review hundreds of systematic reviews and thousands of unique records across various databases. It aimed to comprehensively map out past achievements, current developments, and potential future directions in AI-based medical research. The paper represents the latest brainchild of an open science collaboration project, which welcomes participation from others at https://osf.io/87u6q/. 

The study reveals that as ChatGPT matures, it's finding applications across diverse medical specialties demanding cross-disciplinary collaboration. A fascinating trend shows a transition from theoretical to clinical applications in the AI literature, mirroring developments within the ChatGPT space.

Despite challenges such as ensuring the quality of training data and managing ethical concerns, ChatGPT is fostering the adoption of AI tools within medical areas that have been historically underrepresented in AI application.

If you have an interest in the future of medicine and the role AI technologies like ChatGPT play in shaping it, check out this project. It offers a compelling glimpse into what the future of healthcare might look like. 

REFERENCES

Gabashvili I.S. Artificial Intelligence in Biomedicine: Systematic Review. medRxiv 2023.07.23.23292672; doi: https://doi.org/10.1101/2023.07.23.23292672

Gabashvili I.S. The impact and applications of ChatGPT: a systematic review of literature reviews. arXiv:2305.18086 [cs.CY] https://doi.org/10.48550/arXiv.2305.18086

Gabashvili IS. ChatGPT in Dermatology: A Comprehensive Systematic Review JMIR Preprints. 02/06/2023:48979. medRxiv 2023.06.11.23291252; https://doi.org/10.1101/2023.06.11.23291252

Accelerating Knowledge Innovation: Systematic Review of Reviews on ChatGPT

In the beginning was the Word. It was used in the creation of all other things - from thoughts to stories and histories. However, as the amount of information available grew overwhelming, it became challenging to process and make sense of it all. Early academic reformers introduced the idea of reviews and digests to help navigate this sea of information. But it wasn't until the 1970s that systematic reviews gained popularity, starting in the field of medical research.

created by Author with ChatGPT, Bing Image Creator & Photoshop

Reviews play a crucial role in consolidating a vast array of studies and publications, allowing researchers to weave together the threads of evidence and create a cohesive and informative narrative. However, traditional review processes often require substantial time and the collaboration of multiple researchers to reach a consensus. With the rise of generative AI based on large language models, the power of words can be harnessed to streamline the systematic review process, unlocking new possibilities for learning and knowledge acquisition. By leveraging the capabilities of ChatGPT, researchers can potentially accelerate the production of high-quality reviews, facilitating the dissemination of insights and advancements in various fields of study.

The use of ChatGPT in conducting a systematic review of reviews on ChatGPT itself demonstrates the potential for accelerating the production of high-quality reviews in a timely manner. 

While systematic reviews are considered to be the gold standard in knowledge synthesis, they usually require between 6 months and 2 years to complete and often have a narrow focus. While the methodological shortcuts allow rapid reviews (first mentioned in the literature in 1997) to be conducted in less time and with fewer resources, they also increase the likelihood of introducing bias into the review process and missing important information from grey literature (i.e., preprint servers). 

In 2020, full systematic review was completed by a team of 6 in 2 weeks using automation tools. The most time-consuming tasks were data extraction, write-up, abstract screening, full-text screening, and risk of bias. 4 out of the 6 people on the team were experienced systematic reviewers with complementary skills (three experts in two domains required for the review and one information specialist). 

In 2023, ChatGPT and I were able to complete the review of reviews in one week. We screened 7 large resources of papers, including grey literature and reviewed primary studies in Chinese, German, Indonesian, Norwegian, Portuguese, Russian, and Spanish, in addition to English. 

ChatGPT helped me to filter relevant literature in all languages, extract key information, summarize findings, and even assisted with the synthesis of the overall review, enabling a more efficient and comprehensive analysis.

Our paper illustrates that ChatGPT is expanding into different domains and highlights the need to continually refine and expand the training datasets, ensuring that they are diverse and accurate. Another area of improvement involves developing customized integrations, designing specialized prompt instructions and involvement of domain-specific expert trainers, factual correctness evaluation, and investigation of societal impact.  

Word by word, paper by paper, and review by review, ChatGPT is paving the way for a future where knowledge creation is accelerated, insights are amplified, and breakthroughs are within closer reach.

REFERENCE

Gabashvili I.S. The impact and applications of ChatGPT: a systematic review of literature reviews. Submitted on May 8, 2023. arXiv:2305.18086 [cs.CY]. https://doi.org/10.48550/arXiv.2305.18086

Depression AI

Wearable AI is a promising tool for depression detection and prediction although it is in its infancy and not yet ready for use in clinical practice - as concluded in a recent review. AI can be also used therapeutically. 

Research has shown that interacting with technology, such as chatbots, can lead to feelings of social connection and companionship, which can have both positive and negative effects on mental well-being. Chatbots have become increasingly popular in mental health domain because of their impact on social interactions and the ability to form and maintain meaningful relationships. They are effective in reducing symptoms of anxiety and depression, although there is always a risk that they may exacerbate mental health issues. 

One of the main benefits of chatbots is their ability to provide low-cost and easily accessible mental health counseling. ChatGPT studies show that its potential for depression detection and treatment should be further explored, while addressing the challenges and ethical considerations. ChatGPT outperforms traditional neural network methods but still has a significant gap with advanced task-specific methods. 

In the US, one in five individuals is affected by mental health issues each year, with recreational cannabis use increasing the risk. Intelligent wearables utilize over 30 types of data to predict depression, with physical activity, sleep, heart rate, and mental health measures being the most commonly used. The Depresjon dataset (motor activity recordings of 23 unipolar and bipolar depressed patients and 32 healthy controls) is most popular among researchers.

Previous systematic reviews have shown that AI has better performance in detecting patients without depression than those with depression, but the review published last week shows slightly higher sensitivity and specificity - based on data from wearable devices. It is recommended that tech companies develop wearable devices that can detect and predict depression in real-time. Neuroimaging data in addition to wearable devices would provide even higher diagnostic performance.

With the increasing popularity of IoT and AI, they will likely become an integral part of our lives.  It may soon become a useful tool in clinical practice.

REFERENCES

Abd-Alrazaq A, AlSaad R, Shuweihdi F, Ahmed A, Aziz S, Sheikh J. Systematic review and meta-analysis of performance of wearable artificial intelligence in detecting and predicting depression. NPJ Digit Med. 2023 May 5;6(1):84. doi: 10.1038/s41746-023-00828-5. PMID: 37147384.

Garcia-Ceja E, Riegler M, Jakobsen P, Tørresen J, Nordgreen T, Oedegaard KJ, Fasmer OB. Depresjon: a motor activity database of depression episodes in unipolar and bipolar patients. In Proceedings of the 9th ACM multimedia systems conference 2018 Jun 12 (pp. 472-477).

Lamichhane B. Evaluation of ChatGPT for NLP-based Mental Health Applications. arXiv preprint arXiv:2303.15727. 2023 Mar 28. 

Yang K, Ji S, Zhang T, Xie Q, Ananiadou S. On the Evaluations of ChatGPT and Emotion-enhanced Prompting for Mental Health Analysis. arXiv preprint arXiv:2304.03347. 2023 Apr 6.

Dana RA, Gavril RA. Exploring the psychological implications of ChatGPT: a qualitative study. Journal Plus Education. 2023 May 1;32(1):43-55.

AI for Eyes

Ophthalmology is dominated by imaging. Volumetric, three dimensional (3D) ophthalmic imaging using optical coherence tomography (OCT) has revolutionized assessment of the eye and artificial intelligence (AI) improved clinical decision-making.  Current commercial OCT instruments, especially spectral domain (SD) OCT, are widely used in diagnosis and management of patients with retinal diseases.  Yet, standard 2D cross-sectional images of the retina, that remain the most commonly used OCT images and can be even taken by patients themselves, using smartphone apps, can also provide valuable information utilizing AI models. 

Fundus photography - serial photographs of the interior of the eye (opposite the lens)

taken through the pupil by low-power microscope can help to examine optic disc, retina, and lens. With the drastic improvement in smartphone optics, smartphone fundoscopy has been used with increasing frequency since 2010. Machine learning, particularly deep learning, could analyze millions of such images, to identify and quantify pathological features in almost every ophthalmic disease. Even more, it can detect other health conditions such as hypertension, stroke risk, heart disease, and diabetes. 

            

Using deep learning models, systolic blood pressure could be detected as hypertensive with 60% accuracy (Dai t al., 2020) or within 11 mmHg, major cardiac adverse events with accuracy 70% (Poplin et al., 2018) and glaucoma predicted with 96% accuracy (Gheisari et al, 2021). 

The ImageNet dataset - a very large collection of human annotated photographs (over 14 mln)  - is a good starting point for obtaining a model that performs well in recognizing retinal images. A well-known class of deep neural networks - such as a successful CNN trained on ImageNet can be applied to a retinal dataset, and another classifier learns to work with CNN-encoded features - the method known as transfer learning.  Deep learning can be also combined with traditional machine learning methods and fine-tuning approaches. However, many retinal health variables, such as intraocular pressure, cannot be yet adequately predicted from clinical parameters or retinal photographs even using state-of-art molecular learning or deep learning techniques (Ishii et al., 2021). Only one out of three AI-based algorithms designed to detect diabetic retinopathy was able to outperform human screeners. Possibly, we just need more data. But we might be also needing new models.  

Three principal applications of AI for image analysis are classification, segmentation and prediction.  Automated image segmentation and classification can be done without AI methods, just by  applying a set of mathematical functions on the content of an image and classic ML approaches like SVM or random forest. Deep learning approaches could enhance these tasks. One of newer deep learning techniques, generative adversarial network (GAN), can greatly improve resolution of images (super resolution (SR) estimation from a low-resolution counterpart) and image segmentation. GANs can be also used to synthesize images with various eye pathologies, increasing accuracy of classification tasks. 

Thanks to the advances in AI and smart portable or home devices, the future of medicine, including teleophthalmology, is truly exciting. 

REFERENCES

Schmidt-Erfurth U, Sadeghipour A, Gerendas BS, Waldstein SM, Bogunović H. Artificial intelligence in retina. Progress in retinal and eye research. 2018 Nov 1;67:1-29.  

Gheisari S, Shariflou S, Phu J, Kennedy PJ, Agar A, Kalloniatis M, Golzan SM. A combined convolutional and recurrent neural network for enhanced glaucoma detection. Scientific reports. 2021 Jan 21;11(1):1-1.

Poplin R, Varadarajan AV, Blumer K, Liu Y, McConnell MV, Corrado GS, Peng L, Webster DR. Prediction of cardiovascular risk factors from retinal fundus photographs via deep learning. Nature Biomedical Engineering. 2018 Mar;2(3):158-64.

Dai G, He W, Xu L, Pazo EE, Lin T, Liu S, Zhang C. Exploring the effect of hypertension on retinal microvasculature using deep learning on East Asian population. PloS one. 2020 Mar 5;15(3):e0230111.

Ishii K, Asaoka R, Omoto T, Mitaki S, Fujino Y, Murata H, Onoda K, Nagai A, Yamaguchi S, Obana A, Tanito M. Predicting intraocular pressure using systemic variables or fundus photography with deep learning in a health examination cohort. Scientific Reports. 2021 Feb 11;11(1):1-0.

Sorry I did not quite get that, try again

The holy grail of AI is to fully understand human language in all its nuances. To do that, it should be able to assess, extract and evaluate information from textual data. Were are we now in 2020? 

Do You Want AI with that?

When you hear about Artificial Intelligence, you may picture Ex Machina, Hal 2001, Siri or Alexa.  You may also recall flashy news headlines about AI predicting specific health events and outcomes more accurately than a doctor, navigating better than humans, outperforming government workers, bankers, trial jurors and psychologists. 

 But there are many other new ideas out there. This year, Y combinator's online startup school ...

A Future without Employment

Artificial intelligence has reached a buzzword utopia as we are getting ready for self-driving cars, delivery drones and virtual assistants with human-level intelligence. 

Many believe that this new era of AI  will enable a new kind of American Dream - an early retirement in a country cabin with home grown vegetables and beautiful nature settings. But many others are concerned about a greater inequality created by the jobless future. ..