Healthcare AI Guy Weekly | 5/21

Interested in AI in healthcare? Join 5000+ readers (UHG, a16z, Google, etc)

Want to get in front of healthcare AI enthusiasts? Respond to this email!

Our Picks ✨

Highlights if you’ve only got 2 minutes…

1/

New multimodal AI from Google

Last week, a day after OpenAI’s new product release, Google had its I/O Developer’s Conference. A lot of challengers are coming for Google’s perceived AI throne but they announced their own set of impressive updates, most importantly Project Astra.

Google's Project Astra, a real-time, multimodal AI assistant, is “the future of AI at Google,” according to Demis Hassabis, the head of Google DeepMind. Astra can identify objects, answer questions, and assist with tasks in a conversational manner. The project is part of Google's Gemini announcements, which include new models like Gemini 1.5 Flash for faster task completion, Veo for video generation from text prompts, and Gemini Nano for local device use. Hassabis envisions AI's future to be less about the models and more about their functionality, with AI agents performing tasks on behalf of users. He believes that Astra is closer to a true real-time AI assistant and that improving speed and latency has been a key focus for the team.

With real-time multimodal AI, like Astra or ChatGPTs ‘GPT-4o’, we’ll get better and better applications (b/c the foundation models are better). So for healthcare one could envision an even better and more personalized companion or a personal health assistant who can use computer vision to identify a rash or injury and recommend next steps. (link)

2/

Using AI to reduce administrative workload has limits

Several studies done at academic hospitals have revealed significant limitations of large language models in medical settings. These contrast some of the common industry talking points that AI will always save time, money, and reduce work. Here’s the breakdown:

• A University of California, San Diego study “found that use of an LLM to reply to patient messages did not save clinicians time.” (link)

• A Mount Sinai study “found that popular LLMs are lousy at mapping patients’ illnesses to diagnostic codes.” (link)

• A Mass General Brigham study “found that an LLM made safety errors in responding to simulated questions from cancer patients. One reply was potentially lethal.” (link)

The studies were not universally negative, nor were they pessimistic about the long-term prospects of AI. I think we all know it will take some time for AI to work on the level we expect in healthcare, given the high-stakes nature and industry regulations. That said, in this beginning stage it’s good to know how the technology works and understand current limitations so it can work safely and effectively going forward. (link)

3/

The generative AI roadmap

The IHI Lucian Leape Institute, a think tank within the Institute for Healthcare Improvement, brought together an expert panel to build out a report on AI in healthcare. The panel included Amazon, Google, Microsoft, Harvard Medical School, Kaiser Permanente and others. The goal was to identify the main areas where generative AI is likely to be used as well as challenges and mitigations. Here are the highlights:

• Main benefits: GenAI can reduce clinician burnout, improve diagnostic accuracy, and cut the cost of care.

• Key risks: Potential depersonalization of care, inaccuracies and bias, challenges with integration and workforce deskilling.

• Mitigation strategies: Best practices include serving and safeguarding patients, engaging with clinicians, ensuring AI efficacy and freedom from bias, establishing strict governance, and promoting collaborative learning across health systems.

• Three primary use cases for generative AI: documentation support, clinical decision support, and patient chatbots, each with its own set of benefits and challenges.

While AI holds a lot of promise in transforming healthcare (alleviating burdens), concerns around transparency, bias and clinical deskilling remain — all of which come as no real surprise. (link)

Tools & Partnerships 🔧

Latest on business, consumer, and clinical healthcare AI tools and partnerships…

TOOLS

• Apple and Google using AI to create accessibility features: Apple announced new accessibility features coming to iOS 18, including AI-powered Eye Tracking, Music Haptics, Vocal Shortcuts, and more. Google announced updated AI accessibility features, on Android devices, designed to help people with mobility, vision, or speech impairments. (apple)(google)

• Elon Musk’s Neuralink gets FDA approval for more implants: Neuralink, Elon Musk’s brain implant startup, is accepting applications for a second participant to test out its “Telepathy” implant. The company wants to embed chips in 11 people this year, and reach 22k+ brains by 2030. (link)

• Transcarent new AI-enabled platform ‘WayFinding’: Transcarent, Glen Tullman new telehealth startup, is teasing a new AI chatbot built on ChatGPT. The device will aim to answer the health insurance-related questions that regularly stump more than half of Americans, like: How much will I have to pay for this doctor’s visit? What’s my deductible? Can you help me find a doctor? (link)

• AI-powered GLP-1 data library: Dandelion Health, a real-world data and clinical AI platform, has launched a GLP-1 data library as a multimodal real-world clinical data set built specifically to surface insights and opportunities related to the GLP-1 receptor agonist drug class. The company collected structured and unstructured data for more than 10 million patients across a range of populations and longitudinal patient journeys. (link)

PARTNERSHIPS

• Banner Health + Regard: Banner Health, based in Arizona, is giving clinicians in all 33 of its hospitals across six states access to an AI tool within the EHR that summarizes clinical notes. Developed by Regard, the technology is designed to reduce the clinician’s time spent in front of a computer and facilitate easier access to decision support for care management. (link)

• Nabla + Stratum Med: Nabla inked a new partnership with Stratum Med, bringing ambient AI within reach of over 12k physicians across a hundred medical groups. Stratum will be hands-on with the partnership to not only streamline documentation workflows for providers, but also to coordinate steering committees to drive product innovation, enable new feature validation through targeted pilot programs, and inform specialty template optimization by facilitating clinician feedback loops. (link)

Deal Desk 💸 

Spotlight on latest capital raises, M&A, and investments…

FUNDING

• SmarterDX, a startup helping hospitals automate their medical coding process and collect more on claims, raised $50M in Series B funding led by Transformation Capital. The round also includes continued investments from Bessemer Venture Partners, Flare Capital Partners and Floodgate Fund. (link)

• Crosby Health, a NYC-based developer of AI technology designed to automate administrative tasks for hospitals, raised $2.2M in pre-seed funding. Amplo Ventures led the round and was joined by NOMO Ventures and angel investors. (link)

• Hexis, an AI-powered personalized sports nutrition app, raised $2M in pre-seed funding round co-led by Apex Capital and Sport Republic. (link)

INVESTMENTS

• Accenture + Turbine, Accenture (NYSE: ACN) made a strategic investment, through Accenture Ventures, in Turbine, a predictive simulation company that is building a platform for interpreting human biology. Accenture’s investment will help Turbine further extend its capabilities to global biopharma companies that can benefit from Turbine’s ability to uncover hidden biological insights, with the potential to guide and accelerate key drug development workstreams. (link)

Other Relevant News 🔍

News, podcasts, blogs, tweets, resources, etc…

• Senators release framework for AI legislative action (link)

• Office of National Coordinator for Health IT to provide AI grants (link)

• LLMs could help assess patient acuity in emergency departments (link)

• Report: health systems implementing AI should have strict oversight (link)

• AI not yet seeing wide adoption by physicians (link)

• How AI can bring doctors, patients closer together (link)

Visuals of the Week 📸

Funny memes, cool pics, and interesting data from around the web…