<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Nihar Ganju MD on Medium]]></title>
        <description><![CDATA[Stories by Nihar Ganju MD on Medium]]></description>
        <link>https://medium.com/@niharganju?source=rss-2b780ca92eaf------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*asvkHo3QXBDYcC8uRGJwDQ.jpeg</url>
            <title>Stories by Nihar Ganju MD on Medium</title>
            <link>https://medium.com/@niharganju?source=rss-2b780ca92eaf------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Mon, 16 Nov 2026 08:39:06 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@niharganju/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Modern Virtual Practice: Upgraded with AI]]></title>
            <link>https://niharganju.medium.com/the-modern-virtual-practice-upgraded-with-ai-59887e3898d4?source=rss-2b780ca92eaf------2</link>
            <guid isPermaLink="false">https://medium.com/p/59887e3898d4</guid>
            <category><![CDATA[emr]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[no-code]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[practice-management]]></category>
            <dc:creator><![CDATA[Nihar Ganju MD]]></dc:creator>
            <pubDate>Sat, 10 Aug 2024 05:40:23 GMT</pubDate>
            <atom:updated>2024-08-14T01:44:58.704Z</atom:updated>
            <content:encoded><![CDATA[<h3>5 ways we use AI and no-code tools to run low-maintenance clinical operations.</h3><p>This year we launched our telemedicine practice, <a href="https://www.jointhecollaborative.com/">The Collaborative</a> — a practice of elite national expert doctors focused exclusively on treating menopause and perimenopause with operations in nearly all 50 states. As co-founder and CTO, my key responsibility was to build the right infrastructure for an entirely virtual medical company, operated by a fully remote workforce and distributed leadership.</p><p>As I’m building in 2024, a distinct advantage I have, not available before, is the availability of AI tools and mature no-code tools. I’ve built three virtual care women’s health practices over the last decade. Building the technology infrastructure for a national telemedicine group is complex and expensive requiring investment from venture capital. This is not ideal, as the VC model doesn’t align well with the direct practice of medicine. So part of this experiment was to see if the time has come to build infrastructure for a national telemedicine group without heavy VC investment. To date, we’ve taken no venture capital investment.</p><p>So in this article, I want to demonstrate a selection of five innovative tools we developed for our environment.</p><h3>1. Real-Time Notifications: The Heartbeat of Our Operations</h3><p>Imagine this: every time a patient books, reschedules, or cancels an appointment, a payment is made, a Slack or Google Chat message arrives to every team member’s channel. We have such a system which functions as a a pulse check for all of our teams. Every time an important financial or operational event occurs real-time notification keeps all of our teams informed.</p><p>The technology team uses it as a way to keep tabs that all our 3rd party integrations are firing smoothly; the marketing team uses it as a way to immediately see performance from their social media campaigns, the C-suite uses it as a way to see that all areas of the company are operating smoothly.</p><p>This real-time flow of information means no more waiting for end-of-day reports or sifting through emails to see what’s happening. Everyone is on the same page, which translates into quicker decision-making and a more responsive practice.</p><h4>How we did it</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/797/1*5VB3Oqwd1-kyo5nK0CR14g@2x.png" /><figcaption>The left ventricle is the name of our channel — a tongue in cheek name for the comforting pulse of our operations running smoothly</figcaption></figure><p>Our team uses Google Chat internally. We have a channel that any interested team member can subscribe to and observe the flow of activity of the virtual operations.</p><h3>2. Self-Managed Scheduling: Empowering Patients and Providers</h3><p>In our practice, patients and providers manage their own schedules seamlessly. We don’t have cumbersome, centralized, one size fits all appointment templates that is common in EMR’s. This means there is no individual responsible for setting appointment templates, opening or closing blocks for changes, to the calendar.</p><p><strong>Patient-Led Scheduling</strong> A patient who suddenly has a conflict can open our app, find a new available slot, and reschedule their appointment in seconds — no phone calls, no fuss. This process not only saves time but also gives patients control over their healthcare, which is a huge satisfier.</p><p><strong>Doctor-Led Schedule Management</strong> Doctors, too, have full control. If a physician needs to close a block of time for a last-minute commitment, they can do so directly in the system. Conversely, if they find an unexpected free hour, they can open it up for patient bookings, filling their day more efficiently.</p><p>This flexibility is key to running a modern practice. It keeps patients happy by making it easy to get the care they need, and it helps providers maintain a balanced, manageable schedule.</p><h3>3. AI-Data Analyst: Daily Operational Reports</h3><p>At the end of each day, we generate a report summarizing key metrics like appointment volumes, patient feedback, and daily revenue. The automation is set up using a flow in <a href="https://www.n8n.com/">n8n</a>. The flow is set up with an AI data analyst flow which not only presents the numbers, but compares trends, and for instance, if the AI detects a drop in appointment bookings on a particular day, it might suggest this could be due to a recent holiday or concurrent events. We continue to work on improving the AI analytical ability by giving it access to additional data sources that it can use to perform additional analysis and generate insights useful to our operational team. A short clip of the summary is then sent to our notification channel.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/785/1*8t-uuOuMv3-F7RBYHJHjVA@2x.png" /><figcaption>The nightly AI generated summary includes a friendly joke to keep us entertained.</figcaption></figure><p>Every month, our AI tools additionally perform a deep dive into the data, automatically generating Excel files and graphical reports that break down how our scheduling and revenue systems performed. If there’s a trend of more patients rescheduling certain types of appointments, the AI flags this for us to investigate further.</p><p>These AI-generated insights save hours of manual data crunching and give us a clear view of what’s working and what isn’t. We can quickly pivot our strategies based on real data, rather than guesswork.</p><h3>4. Internal tools built with no-code and AI</h3><p>Our tools we use were developed in-house, tailored specifically for the needs of our practice. This gives us an edge — our systems aren’t generic; they’re designed to meet the exact demands of our virtual operation.</p><p>We’ve extensively used n8n in a self-hosted environment to ensure privacy and HIPAA compliance. This platform allows us to quickly connect our database to automation and AI processes. We use it to populate Google Sheets for reporting and send real-time notifications via Google Chat or Slack, keeping our team informed and aligned.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XBUYDaLJ335dC8NCk-Sg6g@2x.png" /><figcaption>An example of our n8n flow that retrieves data from a Google Sheets file, processes it using OpenAI and communicates the data to our real-time notification channel</figcaption></figure><h3>5. Lead Generation with AI Micro-tools</h3><p>We’re discovering that AI micro-tools are a potent lead generation strategy. We deployed multiple AI micro-tools that provide valuable education or information personalized to each user. For example, we started by offering targeted quizzes on topics such as the <a href="https://www.realdocai.com/mud-voice-0175">optimal vaginal cream for menopausal symptoms</a> and the specific stage of perimenopause or menopause a user might be experiencing. We engage potential clients in a personalized, informative journey. This interactive approach helps users foster a sense of trust and relevance, guiding users through our offerings in a manner that is both educational and customized to their needs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/560/1*gN_kEqJzAoznDv9oM015LQ@2x.png" /><figcaption>The AI micro-tools offer our users a brief, anonymous but personalized engagement with our chatbot.</figcaption></figure><p>The quizzes are anonymous but helps us understand our user’s unique circumstances and preferences, allowing for tailored recommendations. Moreover, the data gleaned from these quizzes is invaluable in refining our marketing strategies and enhancing our service offerings.</p><p>By directing social media and Google ad traffic to these microtools, we not only increase user engagement but also seamlessly transition interested individuals from casual browsers to potential clients, thus optimizing our conversion rates and driving growth in our virtual practice.</p><h3>Final Thoughts</h3><p>Running a virtual practice today means embracing these innovations fully — because in 2024, technology isn’t just part of the practice; it is the practice.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=59887e3898d4" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Alignment in Medicine. Clinical Trust and Risk of Institutional Bias]]></title>
            <link>https://niharganju.medium.com/ai-alignment-in-medicine-clinical-trust-and-risk-of-institutional-bias-d5499d763379?source=rss-2b780ca92eaf------2</link>
            <guid isPermaLink="false">https://medium.com/p/d5499d763379</guid>
            <category><![CDATA[medical-ai]]></category>
            <category><![CDATA[digital-health]]></category>
            <category><![CDATA[llm]]></category>
            <dc:creator><![CDATA[Nihar Ganju MD]]></dc:creator>
            <pubDate>Wed, 06 Mar 2024 17:41:32 GMT</pubDate>
            <atom:updated>2024-03-13T17:31:21.477Z</atom:updated>
            <content:encoded><![CDATA[<h3>Alignment in Medical AI. Clinical Safety and Risk of Institutional Bias</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*P4H_3HGkVlFrHKFc8Zcgbw.gif" /></figure><h3>Introduction</h3><p>As my team continues to operate on the frontiers of Artificial Intelligence (AI) and healthcare, the question of AI alignment continues to remain an alarming one. The crux of AI’s integration into healthcare hinges on <em>alignment</em> — a term that emphasizes the necessity for AI systems to act in accordance with human values and ethics, especially in sensitive sectors like healthcare. The concept of AI alignment in healthcare needs to be revisited repeatedly by those of us operating in the field to answer the pivotal question: How do we achieve medical alignment, and whom do we entrust with this critical responsibility?</p><h3>Medically Aligned AI</h3><p>Discussions with my physician peers, who are developing AI in healthcare, we are settling on the concept of medically aligned AI — a specialized approach to AI models that emphasizes personalization and physician attribution.</p><p>Unlike current AI development in healthcare applications, medically aligned AI advocates for personalized AI training, such that a unique model is developed for each physician. This goes beyond the current training cycles which are merely overseen with guidance of anonymous doctors. This method of alignment ensures that every piece of advice or diagnosis the model provides is directly attributed to the individual physicians. This results in not one monolithic medical model for all, but many models for many doctors.</p><p>This idea starkly contrasts with other models of alignment where the focus might inadvertently shift from patient care to other institutional stakeholders, including corporate, government or medical institutions, whose primary objectives may not align with the personalized needs of the patient.</p><h3>Framework for Developing Medically Aligned AI</h3><p>Medically aligned AI is defined by two core principles: the publicly published involvement of licensed medical professionals in model training and the direct attribution of its outputs to individual physicians.</p><p>While medical institutions play a crucial role in the dissemination of knowledge and the training of healthcare professionals, their contribution lacks the personal touch and direct accountability inherent in physician-patient relationships. On the other end of the spectrum, corporate-directed AI training, influenced by the imperatives of profit and market share, often misaligns with the fundamental goal of patient care, treating individuals as consumers rather than patients. Likewise, government coerced medical practice has become all too real to those of us practicing women’s health, and AI physician researchers are cautious and wary of this influence embedded in generic publicly available AI models.</p><h3>Personalization Creates Transparency and Trust</h3><p>The medically aligned AI principles advocates for physician personalization rather than institutional personalization. Personalized models allow a physician’s to be paired with their AI model. This personalization also allows personality mimicking by the AI. The seamless extension enables a believable and reliable model extending trust of the patient-physician relationship. Personalization also creates a deeply human-centric AI model which accepts that variations of opinion of each physician not only exist, but are healthy and desirable. It acknowledges and decidedly leans into the idea that the human physician’s are inherently variable and can be opinionated. Their interpretation of evidence based literature combined with experience driven decisions are not only appropriate, it is desired and improves the human connection patients seek.</p><p>Patients are more likely to trust and follow medical advice when they know it originates from their doctor, rather than an impersonal algorithm or institution. This direct attribution not only enhances the credibility of AI-powered recommendations but also strengthens the bond between patients and their physicians.</p><h3>In Practice: Bringing Theory to Life</h3><p>Real-world applications of medically aligned AI, though in nascent stages, offer a glimpse into successful personalization that can be achieved at scale. Several teams, including ours, have developed the technology for rapid cycle customized AI models for physicians. Our soon to be published data from our team and others is imminent on the ability to create usable, accurate, and attributable and personalized AI models.</p><p>From AI systems that accurately predict patient deterioration hours before conventional methods to algorithms that tailor rehabilitation exercises to the specific recovery trajectories of patients, the potential is immense. These examples not only validate the feasibility of medically aligned AI but also highlight its profound impact on patient care and outcomes.</p><h3>Conclusion</h3><h4>Medically Aligned AI</h4><p>Aligning AI with human values in medicine must prioritize personalization and physician attribution, ensuring outputs are attributed to specific physicians, enhancing personal accountability and patient trust.</p><h4>Personalization and Trust</h4><p>Leverage physician-specific AI models to mirror personal and professional characteristics, fostering a stronger patient-physician relationship through trusted, personalized AI advice.</p><h4>Achievable?</h4><p>Yes! Technology is now available which can successfully train personalized AI models for physicians.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d5499d763379" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Medical-AI: How should we evaluate performance in medical AI models?]]></title>
            <link>https://niharganju.medium.com/medical-ai-how-should-we-evaluate-performance-in-medical-ai-models-639cf71b3218?source=rss-2b780ca92eaf------2</link>
            <guid isPermaLink="false">https://medium.com/p/639cf71b3218</guid>
            <category><![CDATA[menopause]]></category>
            <category><![CDATA[digital-health]]></category>
            <category><![CDATA[medical-ai]]></category>
            <category><![CDATA[llm]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Nihar Ganju MD]]></dc:creator>
            <pubDate>Thu, 05 Oct 2023 06:10:17 GMT</pubDate>
            <atom:updated>2023-10-05T06:10:17.569Z</atom:updated>
            <content:encoded><![CDATA[<h4>We developed a specialized medical AI model purpose-built for women’s health. But how should medical models be evaluated? Here’s what we did: Real questions, from real patients, evaluated by real doctors.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*0trFEAoEyfJCEB9F" /><figcaption>Photo by <a href="https://unsplash.com/@towfiqu999999?utm_source=medium&amp;utm_medium=referral">Towfiqu barbhuiya</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h3>Benchmarks to real-world relevance</h3><p>I’ve been looking at data for medical benchmarks for AI models in the field of medicine. Examples include Med-PaLM[<a href="https://arxiv.org/abs/2212.13138">1</a>][<a href="https://arxiv.org/pdf/2305.09617.pdf">2</a>], <a href="https://www.medrxiv.org/content/10.1101/2022.12.19.22283643v2.full">GPT3.5</a>, <a href="https://arxiv.org/pdf/2303.13375.pdf">GPT4</a> and <a href="https://www.hippocraticai.com/benchmarks">HippocraticAI</a>. These models all assessed their performance with simulated medical board questions.</p><p>Here is a sample of the published performance metrics on USMLE exams:</p><p><a href="https://arxiv.org/abs/2212.13138">MedPaLM</a>: 67.2% accuracy (Singhal, Dec 2022)<br><a href="https://arxiv.org/pdf/2305.09617.pdf">MedPaLM2</a>: 86.5% (Singhal, May 2023)<br><a href="https://www.medrxiv.org/content/10.1101/2022.12.19.22283643v2.full">GPT-3.5</a>: ~60+% (Kung, Dec 2022)<br><a href="https://arxiv.org/pdf/2303.13375.pdf">GPT-4</a>: ~86+% (Nori, Apr 2023)<br><a href="https://www.hippocraticai.com/benchmarks">HippocraticAI</a>: 89.9 (Published on marketing site, May 2023)</p><p>While these performances make for fun headlines like <a href="https://gizmodo.com/ai-chatgpt-bing-google-passed-step-1-medical-exam-1850109825"><em>GPT passes the Step 1 medical exam</em></a><em>, </em>the benchmarks remain academic demonstrations of the technology. The industry appears to be moving towards demonstrating ever increasing ability to ace multiple medical exams — but what does that really mean for clinicians? From a clinicians’ perspective, these technology evaluations don’t <em>yet</em> offer any practical clinical take-aways that substantially advances medicine or medical training.</p><h3>The AI model for women’s health</h3><p>Our team has spent the last several months developing several AI models capable of chat-based question answering in the women’s health domain. As the models improved, it was time to clinically evaluate our menopause AI model first (others to come later).</p><h3>Real Patients, Real Questions</h3><p>We could have used the NAMS <a href="https://www.menopause.org/for-professionals/ncmp-certification">Menopause certification</a> question bank to evaluate the AI model. But, as doctors, we do have the privilege of engaging patients daily in multiple settings. We took advantage of this. We asked followers of a menopause centric social media channel to anonymously share their real questions and health challenges.</p><p>Then, we selected 100 questions from the submission pool, asked the chat model the questions, and captured the output as model generated responses.</p><p>What we see in these questions is that patients don’t ask questions like board exams. These questions are raw, unstructured and diverse. And they certainly don’t give you a multiple choice set of answers to select from.</p><h3>Evaluation by Real MD’s</h3><p>We created a document of 100 question/answer pairs similar to this sample:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KpJDGG1Q9t8Q0Td0Scq-Cw.png" /><figcaption>Real questions, AI generated responses. These question/answer pairs were evaluated by MD menopause experts for accuracy and quality.</figcaption></figure><p>Then, we recruited MD’s who practice menopause as their core speciality. These MD experts evaluated the AI generated responses for accuracy and quality on a likert scale.</p><h3>The results</h3><p>The menopause model described here achieved a accuracy and quality rating of greater than <strong>80%</strong>. We are proud of this. More significantly, we introduced a practical and clinically relevant method for evaluating AI models. By using a genuine real-world question dataset, assessed by experienced doctors, we established a realistic benchmark of clinical accuracy. This approach allows a doctor to truly grasp the capabilities of a new AI model and instills trust, especially when assessed by experts in the field.</p><p>Summarized Key points:<br>✨ Real-World Relevance: Existing benchmarks from GPT-4, Med-PaLM2, Hippocratic AI rely on simulated board exam questions. Instead we used real questions, by real women.<br>✨ Highly accurate model: the model performance matches or outperforms industry benchmarks from GPT-4, Med-PaLM2, HippocraticAI.<br>✨ Validation, not Competition: We think of this not necessarily as “better than” the work from OpenAI and Google, rather this is another validation of the technology.</p><p>We <a href="https://docsend.com/view/3b3n3nicx3u5a46y">presented</a> the study described above as an oral <a href="https://menopause.org/docs/default-source/agm/nams23_regular-lb_abstracts.pdf">abstract</a> at the 2023 NAMS conference.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=639cf71b3218" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I spoke with 10 VC investors in healthcare-AI. These are the takeaways.]]></title>
            <link>https://niharganju.medium.com/i-spoke-with-10-vc-investors-in-healthcare-ai-these-are-the-takeaways-74681a47dd72?source=rss-2b780ca92eaf------2</link>
            <guid isPermaLink="false">https://medium.com/p/74681a47dd72</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[healthcare-ai]]></category>
            <category><![CDATA[medical-ai]]></category>
            <category><![CDATA[venture-capital]]></category>
            <dc:creator><![CDATA[Nihar Ganju MD]]></dc:creator>
            <pubDate>Fri, 14 Jul 2023 18:16:50 GMT</pubDate>
            <atom:updated>2023-07-14T18:25:44.062Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*wLZWROU0Bes3rgLG" /><figcaption>Photo by <a href="https://unsplash.com/@picsbyjameslee?utm_source=medium&amp;utm_medium=referral">James Lee</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h4>Conversations Navigating the Medical-AI Investment Landscape: What worries investors.</h4><p>For the past three months I’ve talked with a variety of tech and health care investors from Silicon Valley, New York, and Europe. My goal was to understand the pulse and worries of investing in medical-AI in the face of the <em>most</em> transformative technology humans have ever developed.</p><p>I’ve summarized the themes below. The perspectives and apprehensions are from 10 leading VC’s and AI experts. I’ve also shared my take on each topic.</p><h4>The Fear of AI Safety in Healthcare</h4><p>Among the VCs and investors I spoke with, safety is a consistent concern. Even while they recognize the immense potential of AI in improving patient outcomes and streamlining healthcare processes, there is a palpable fear of unintended consequences that could arise from AI deployment in medicine, and the subsequent risk to their investment. The potential for misdiagnosis, algorithmic biases, and privacy breaches weighed heavily. Many felt the safest investable path is to fund healthcare AI for non-medical products, thereby avoiding the most challenging aspects of the issue altogether for now. There is comfort in applying AI to back-office tasks, dictation and documentation tasks.</p><p><strong>My Take:</strong> AI for backoffice and dictation tasks may be safer in the early stage but these uses are not the solving the true bottlenecks of healthcare. You don’t even need AI to solve these problems. Cognitively complex tasks are the best suited use-cases for the newest AI and LLM models. It should also be encouraging that <a href="https://arxiv.org/abs/2307.03718">frameworks</a> for regulatory guidance are rapidly emerging — just in time. My conversations with legal experts in the area also suggests that highly regulated industries like healthcare have preexisting principles which are broad enough in their existing forms with the requisite incentives and disincentives to ensure safe AI applications even in the absence of brand new regulations.</p><h4>Worrying about Hallucinations of AI</h4><p>The next common theme was the apprehension around the possibility of AI hallucinations — the idea that AI systems could generate false or misleading information, leading to disastrous consequences in critical decision-making scenarios. Clear regulations, rigorous testing, and continuous monitoring were cited as essential components to mitigate this risk and enhance the reliability of AI systems.</p><p><strong>My take: </strong>There isn’t much research published on this topic specifically but the sense in the research community is that this may be due to limitations from the architecture of the neural networks architecture not yet designed to understand the negative or inhibitory signals. I’ve learned from our prototypes developed at our company (currently in stealth), methods that are quite effective to reduce hallucination risk significantly. We’ve also explored AI architecture designed to present its references and source material. This appears to be a solvable problem.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*HtLNP554omqBlXLh" /><figcaption>Photo by <a href="https://unsplash.com/@iammottakin?utm_source=medium&amp;utm_medium=referral">Mojahid Mottakin</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><h4>Cautious Embrace of AI</h4><p>Despite the investments pouring into AI companies, VCs and investors admitted they are cautious of fully embracing AI. While they may have an investment thesis for AI they admit that they struggle with understanding the <em>technical</em> limitations of AI given its complexity. The 10 I spoke with admitted they have not read a single journal article published in a scholarly or peer-reviewed journal. Instead they have to rely primarily on expert opinions of their peer network and summaries in business publications.</p><p>This is not surprising. The field is complex. The mathematics is not straightforward. And there are a limited number of experts. Then layer on the complexity of applying this to the healthcare world. Thus from these investors perspective it becomes hard to quantify risk.</p><p><strong>My take:</strong> Anyone building, investing, or regulating the AI industry would be wise to start with the first principles of AI. Do the fundamental concepts of computing allow for cognitively capable machines? Review the mathematics. Is it sound? Is mimicking biologic principles for machine development sensible? Track the rate of scholarly research. Are they demonstrating <a href="https://www.youtube.com/watch?v=R0wBMDoFkP0">massive breakthroughs</a>? Is hardware keeping pace with compute power for large model deployment? What is the lifecycle of translation from research to production-grade deployment? It is now in the order of months — not years! We haven’t seen this type of rapid pace in computing and capability before!</p><h4>AI Investing in regulated industries</h4><p>VC’s and Investors in medical-AI are both excited <em>and</em> cautious. Caution is logical. But what indicators offset the apprehension? I argue the central leading indicator in AI for financial forecasting is to monitor the pace of research, given that AI development is moving so fast from research to production. Secondly, highly regulated industries may ironically lower investment risk. The rules and laws of healthcare, particularly regulations of medical practice developed around medical licenses define scope of work (ie. MD, vs RN, vs MA). This actually give us useful parameters by which to architecture AI products.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=74681a47dd72" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>