Imran Qureshi’s Post

We’re #hiring at b.well. Know anyone who might be interested in making healthcare easier for everyone?

There is a lot of hype around Generative AI (LLMs) in Healthcare and Life Sciences. But how do we separate the real world use cases being deployed right now from pie-in-sky dreams of the future? We have 1.2 million physicians in the country but 330 million patients and caregivers. What use cases are being pursued to empower the 330 million people in healthcare using GenAI? We're also planning to delve into some controversial topics such as what current investments in GenAI in healthcare are likely to fail and how the jobs of healthcare workers will change in the next five years. Finally we plan to discuss how we can balance the power and potential of GenAI in healthcare with the ethical concerns and how GenAI can enable patients to be more involved in life sciences. Come join Karim Galil, M.D. from Mendel.ai, Brigham Hyde, Ph.D. from Atropos Health, Samantha Perrone from Amazon Web Services (AWS) and me for a spirited discussion on "How generative AI is transforming healthcare and life sciences." at the AWS Summit in New York on July 10 (12:45 PM-01:30 PM). You can register here: https://lnkd.in/gAzYYzAQ

🚀 𝐓𝐡𝐞 𝐔𝐩𝐜𝐨𝐦𝐢𝐧𝐠 𝐃𝐢𝐬𝐢𝐥𝐥𝐮𝐬𝐢𝐨𝐧𝐦𝐞𝐧𝐭 𝐰𝐢𝐭𝐡 𝐋𝐋𝐌𝐬 🚀 > And why that is a good thing!! Gartner's hype cycle is a proven model for understanding the journey of revolutionary technologies. Today, we're seeing LLM (Large Language Model) technology at the 𝐏𝐞𝐚𝐤 𝐨𝐟 𝐈𝐧𝐟𝐥𝐚𝐭𝐞𝐝 𝐄𝐱𝐩𝐞𝐜𝐭𝐚𝐭𝐢𝐨𝐧𝐬. People expect it to eliminate jobs overnight, solve every problem we have, and be eagerly adopted by everyone. But now we're seeing a move towards the 𝐓𝐫𝐨𝐮𝐠𝐡 𝐨𝐟 𝐃𝐢𝐬𝐢𝐥𝐥𝐮𝐬𝐢𝐨𝐧𝐦𝐞𝐧𝐭, where initial excitement wanes as reality sets in. In 2024, only 63% of companies are planning to increase AI investments compared to 93% in 2023. Having worked extensively with LLM technology over the past year, I believe our expectations are inflated beyond what the LLM technology can deliver TODAY. 🔍 Key Obstacles to Corporate Adoption of LLMs: 1. 𝐇𝐚𝐥𝐥𝐮𝐜𝐢𝐧𝐚𝐭𝐢𝐨𝐧𝐬: Inherent to LLMs, especially in complex use cases. We still have to invent technologies to manage them. 2. 𝐃𝐚𝐭𝐚 𝐅𝐨𝐫𝐦𝐚𝐭: Much of the data in companies is in tabular format which LLMs struggle to understand. 3. 𝐑𝐢𝐠𝐢𝐝 𝐖𝐨𝐫𝐤𝐟𝐥𝐨𝐰𝐬: True value comes when LLMs control workflows, not be controlled by the current rigid workflows and user interfaces. 4. 𝐏𝐫𝐨𝐜𝐞𝐬𝐬 𝐑𝐞𝐢𝐦𝐚𝐠𝐢𝐧𝐢𝐧𝐠: Instead of trying to speed up our current processes, we will achieve true benefit when we reimagine our processes leveraging LLMs. 💡 Navigating the LLM hype cycle: As we move through the hype cycle, the key to success for companies will be focusing on simple use cases, evolving their data platforms, changing their workflows to be dynamic and redesigning their processes. This approach will help us through the 𝐒𝐥𝐨𝐩𝐞 𝐨𝐟 𝐄𝐧𝐥𝐢𝐠𝐡𝐭𝐞𝐧𝐦𝐞𝐧𝐭 to eventually reach the 𝐏𝐥𝐚𝐭𝐞𝐚𝐮 𝐨𝐟 𝐏𝐫𝐨𝐝𝐮𝐜𝐭𝐢𝐯𝐢𝐭𝐲, where the real benefits of LLMs will be realized. Eventually we will look back at LLM technology being as transformative to our corporations and our society as the Internet and the mobile phone were. However we can't achieve that until we lower our expectations to be more realistic today. For a deeper dive into navigating the LLM hype cycle: https://lnkd.in/geX9ejPQ #AI #LLM #Technology #Innovation #HypeCycle #Productivity #BusinessTransformation

𝗪𝗵𝘆 𝗛𝗮𝗹𝗹𝘂𝗰𝗶𝗻𝗮𝘁𝗶𝗼𝗻𝘀 𝗖𝗮𝗻 𝗡𝗲𝘃𝗲𝗿 𝗕𝗲 𝗘𝗹𝗶𝗺𝗶𝗻𝗮𝘁𝗲𝗱 𝗶𝗻 𝗟𝗟𝗠𝘀 *But we CAN develop technologies to minimize them.* There are software companies promising hallucination-free LLMs. Unfortunately, this is physically impossible. I’ll explain below why hallucinations are inherent in LLM technology. ### Types of Hallucinations 1. **Data Quality Hallucinations**: These occur when LLMs are trained on incorrect information. While using only trusted sources can mitigate this, it doesn't address the root cause of all hallucinations. 2. **Inherent Hallucinations**: These arise from the way LLMs function, even with 100% accurate data. This is our focus today. ### Understanding the Inherent Hallucinations The core mechanism of LLMs is predicting the next word in a sequence [Read more](https://lnkd.in/gGefAjxW). To do this, the LLM evaluates the context and calculates probabilities for the next word. In English, there are about 40,000 common words. Considering just the last word and the next word results in 1.6 billion combinations (40,000 x 40,000). If we add one more word, the combinations soar to 64 trillion. For a 100-word paragraph, the combinations are astronomical: 1.6 followed by 131 zeros, a number far exceeding the total grains of sand on Earth's beaches or the stars in the observable universe. Calculating this with today's technology is impossible. ### Compression and Information Loss LLMs use a "transformer" in Generative Pretrained Transformer (GPT) to "compress" information (https://lnkd.in/gzjaHJNp). However, this compression inevitably leads to information loss. Current research suggests that lossless compression of these combinations isn't possible. As a result, LLMs operate on partial information, leading to inherent hallucinations. If LLMs could provide always correct answers without hallucinations, it would imply that lossless compression is achievable, which it isn't. ### Scaling LLMs Increasing the size of LLMs doesn't solve the problem. The sheer number of calculations needed makes it physically impossible to capture all information. Moreover, it's unclear if making LLMs larger continues to reduce hallucinations. ### Mitigation Strategies Since we must accept that hallucinations will occur in every LLM, we can focus on minimizing them. Techniques include: - **Eight-Step Framework for Controlling Hallucinations**: [Read more](https://lnkd.in/gWvy4bkE) - **Collaborative LLMs**: Using multiple LLMs to collaborate for accurate answers. [Read more](https://lnkd.in/gpDUiZmm) - **Human Feedback**: Fine-tuning LLMs with human input. By understanding and applying these strategies, we can better manage hallucinations in LLMs, improving their reliability and accuracy.

🐶 𝑰 𝑱𝒖𝒔𝒕 𝑾𝒂𝒏𝒕 𝒕𝒐 𝑩𝒆 𝑻𝒓𝒆𝒂𝒕𝒆𝒅 𝑳𝒊𝒌𝒆 𝒂 𝑫𝒐𝒈! 🐶 (Why does my dog get better healthcare than me...?) Throughout the year, I receive personalized emails and text messages from my vet, reminding me about the care services my dog needs like the one below. These messages are very personalized, include a photo of my dog and dates on when each service is needed. In contrast, I've been with my primary care provider for five years and have NEVER received any proactive communication from them. Sure, the provider will remind me of an upcoming appointment and the health system sends out a mass email before every flu season, but that's about it. There's a misconception in healthcare that patients will only receive personalized attention if they are in a value-based model. This is a prime example of flawed thinking. My provider could have generated significant revenue over the past five years by simply reminding me of normal, preventative care. The b.well platform enables care providers to engage their patients with personalized notifications and care reminders so the providers make more revenue and the patients become healthier. **Can I please be treated like a dog?** 🐾

"The recognition from Newsweek underscores b.well’s momentum in the digital health space. The company recently joined the Amazon Web Services (AWS) Partner Network and closed a $40 million Series C growth round. This follows the announcement of a groundbreaking integration with Samsung, which aims to give millions of consumers the ability to create longitudinal health records; receive personalized, proactive health insights; and conveniently access relevant in-person, digital and virtual providers, all through the Galaxy smartphone." "Today, organizations like Walgreens, Northwell Health, Lee Health and ThedaCare use b.well to put consumers at the heart of every interaction, derive more value from their technology investments and build ecosystems of care — where data, solutions and services are unified on one convenient platform. Built on a growing network of over 60,000 live, consumer-mediated data connections, b.well leverages the industry’s richest set of longitudinal health data to unlock the value of truly personalized care."

𝐓𝐡𝐞 𝐏𝐨𝐰𝐞𝐫 𝐨𝐟 𝐋𝐋𝐌𝐬 𝐭𝐨 𝐈𝐦𝐩𝐫𝐨𝐯𝐞 𝐭𝐡𝐞 𝐏𝐚𝐭𝐢𝐞𝐧𝐭 𝐄𝐱𝐩𝐞𝐫𝐢𝐞𝐧𝐜𝐞 𝐶𝑎𝑛 𝑙𝑎𝑟𝑔𝑒 𝑙𝑎𝑛𝑔𝑢𝑎𝑔𝑒 𝑚𝑜𝑑𝑒𝑙𝑠 (𝐿𝐿𝑀𝑠) 𝑏𝑒 𝑡ℎ𝑒 𝑠𝑜𝑙𝑢𝑡𝑖𝑜𝑛 𝑡𝑜 ℎ𝑒𝑙𝑝 𝑝𝑎𝑡𝑖𝑒𝑛𝑡𝑠 𝑔𝑒𝑡 𝑡ℎ𝑒 𝑐𝑎𝑟𝑒 𝑡ℎ𝑒𝑦 𝑛𝑒𝑒𝑑, 𝑤ℎ𝑒𝑛 𝑡ℎ𝑒𝑦 𝑛𝑒𝑒𝑑 𝑖𝑡? Watch the video demo: https://lnkd.in/gwSkM4Up