Biolinq Files Semiconductor Sensor With FDA to Power AI Healthcare

Biolinq向FDA申报半导体传感器，赋能AI医疗 Biolinq认为，赋能有效的医疗AI需要连续的多分析物硅基生物传感器，并结合极具吸引力的AI分身，以推动实时的行为改变。 Rich Yang介绍了Biolinq突破性的硅基半导体生物传感器，该传感器可直接从皮肤组织液中采集连续、高保真的生物数据。他解释了多分析物追踪（测量葡萄糖、乳酸、苯丙氨酸等蛋白质以及皮质醇）如何促成实时反馈与行为改变。通过将这种高分辨率的居家纵向健康数据与富有吸引力、性格鲜明的AI分身相结合，Biolinq旨在为医疗基础AI模型赋能，并延长人类的健康寿命。 消费科技与医疗健康领域正在加速融合，科技巨头积极向 FDA 递交 510(k) 申请，但医疗健康 AI 基石模型仍面临居家场景下高保真、高分辨率连续健康数据匮乏的瓶颈。 (0:04) Biolinq 已向 FDA 提交了全球首款硅基半导体生物传感器，该设备利用标准 8 英寸晶圆工艺制造，借助纳米级硅材料的高硬度与微阵列技术，无痛采集皮下组织液以连续测量代谢物。 (4:30) 新型微阵列芯片成功实现了葡萄糖与乳酸的单芯片多通道同步监测，且信号互不干扰，有望彻底颠覆传统的运动乳酸监测与体能评估方式。 (11:24) 基于合成生物学与 DNA 受体技术，Biolinq 正在拓展连续蛋白质（苯丙氨酸）监测、压力皮质醇追踪以及病毒抗原被动筛查等前沿应用。 (14:15) 为打通用户依从性与行为改变的“最后一公里”，AI 数字人分身正在结合实时生物传感器数据，通过个性化、拟人化的即时互动推动无摩擦的行为重塑。 (17:20)

Rich Yang, CEO of Biolinq, argues that realizing the promise of personalized healthcare AI requires continuous, multi-analyte biosensors to capture high-fidelity metabolic data outside clinical settings, coupled with conversational AI avatars that turn raw biological signals into sustainable behavior change. Closing Healthcare AI's Longitudinal Data Gap (0:04) While consumer technology companies and medical device developers are increasingly converging through FDA 510(k) filings, Yang asserts that healthcare foundation AI models remain bottlenecked by a lack of continuous, high-resolution data from patients' daily lives. Existing consumer wearables monitor basic parameters like heart rate variability and pulse oximetry, but lack deeper biological context. Yang identifies diabetes management as the model for success, where devices, algorithms, and continuous glucose monitors (CGMs) answer two core user questions: "Am I okay?" and "What do I do?" While continuous biosensors in diabetes boast some of the market's highest valuations and customer lifetime values, Yang notes that expanding this framework to broader health management requires physical hardware capable of detecting biomarkers present at one-thousandth the concentration of glucose. Silicon Micro-Arrays and Interstitial Fluid Sampling (4:30) To capture low-concentration biomarkers, Biolinq developed a semiconductor biosensor platform built on standard 8-inch silicon wafers, which has been submitted for FDA review. Yang explains that at the nanoscale, silicon is harder than both titanium and stainless steel, enabling the fabrication of durable micro-arrays with bio-receptors hundreds of times thinner than human hair. Designed for metabolic conditions like Type 2 diabetes and currently in clinical trials, the wearable rests on the skin without piercing nerve endings or drawing blood. It samples interstitial fluid—which Yang describes as the body's most metabolically active tissue compartment—and is supported by over 1,000,000 hours of clinical validation comparing sensor readings against 15-minute venous blood draws. The device also integrates ambient LED indicators into the user's peripheral vision to provide passive feedback, such as glowing blue when glucose is within the target range. Multiplexed Sensing via Synthetic DNA Receptors (11:24) Biolinq is expanding beyond glucose by multiplexing distinct sensor chemistries onto a single microchip without signal interference. In athletic performance, continuously tracking lactate replaces disruptive earlobe blood draws to identify when mitochondria peak during exercise, a method Yang reports cyclist Sky Christopherson used to break an indoor velodrome record clean. By synthesizing DNA strands into custom bio-receptors, Biolinq has demonstrated continuous sensing for phenylalanine. Yang highlights that tracking this amino acid eliminates manual food logging and helps address sarcopenia (muscle loss) in patients on GLP-1 weight-loss therapies by verifying adequate protein intake. Yang also forecasts broader applications for synthetic bio-receptors, including continuous cortisol tracking for stress monitoring and wearable patches that illuminate upon detecting viral antigens or antibodies during pandemics. Cognitive Restructuring via Conversational AI Avatars (17:20) Transforming continuous data into sustained behavioral modification requires rapid feedback loops that drive cognitive restructuring. Yang explains that immediate, passive alerts—such as a sensor turning yellow after consuming orange juice—allow users to instantly connect cause and effect, prompting prompt adjustments like taking a short walk or drinking water without requiring manual logging. To bridge the "final mile" of user engagement, Biolinq pairs real-time sensor streams with generative AI avatars, such as a coach named Liz backed by celebrity personalities. Yang contends that providing biological data with a relatable voice—capable of delivering firm or humorous real-time nudges when metrics elevate—converts complex biometric signals into actionable, sustained lifestyle changes. Yang emphasizes that extending human healthspan requires open-architecture biosensing platforms that combine hard-tech synthetic biology with engaging software interfaces. However, broader consumer and clinical adoption remains dependent on regulatory outcomes, as Biolinq's metabolic wearable remains in clinical trials and under FDA review.

Biolinq 首席执行官 Rich Yang 指出，医疗人工智能面临的核心瓶颈在于缺乏来自家庭场景的高保真连续纵向生理数据；通过半导体微阵列与合成生物学技术检测微量生物标志物，并结合生成式 AI 形象提供实时行为干预，是实现个性化预防医学并降低医疗成本的关键路径。 医疗AI的数据瓶颈与慢病闭环模式 (0:04) Rich Yang 指出，当前消费级科技巨头正积极向美国食品药品监督管理局（FDA）提交 510(k) 申请，消费电子与医疗级应用的交汇为降低医疗成本创造了历史契机。然而，实现医疗人工智能的全部潜力仍是最艰巨的挑战。尽管 AI 在医学影像领域已能敏锐识别肿瘤和癌症，但通用基础医疗 AI 模型的发展整体依然受限。其核心阻碍在于，高精度的个性化医疗依赖于直接在患者居家环境中采集的高分辨率连续纵向生理数据，而现有的智能手表仅能捕获心率变异性（HRV）或血氧等基础数据，基础医疗 AI 模型严重缺乏深层生物学背景支撑。 在探索成功路径时，Yang 将糖尿病管理视为范本。该领域多年来整合了硬件设备、算法与机器学习，通过解答用户“我是否安好”与“我该做什么”两个核心问题，建立了极具商业价值的追踪与反馈闭环。在过去 15 年间，糖尿病生物可穿戴设备公司的股票长期位列华尔街表现最佳的资产阵营，享有高估值乘数与高客户生命周期价值（LTV）。然而，AI 无法凭空发现生物标志物，必须依托物理传感硬件。虽然葡萄糖是目前主要的消费级标志物，但人体内许多反映深层健康状态的关键指标，其浓度仅为葡萄糖的千分之一，亟待硬件底层技术的突破。 硅基半导体传感器与无痛间质液检测 (4:30) 针对生理数据匮乏的难题，Biolinq 宣布已向 FDA 提交了全球首款半导体生物传感器的审批申请。该项技术由加州大学圣迭戈分校团队研发，利用标准的 8 英寸硅晶圆制造工艺，将数千个微型传感器直接集成在硅基底上。Yang 解释称，在纳米尺度下，硅材料的物理硬度超越了钛金属和不锈钢，因而是制造微型生物传感器的坚固材料。通过半导体工艺与合成生物学的融合，研发团队制造出厚度仅有人类头发丝几百分之一的生物受体。 该技术平台摒弃了传统电化学传感器深入组织或抽取血液的做法，转而采用紧贴于皮肤表层的微阵列微芯片。微芯片在不触及神经末梢的前提下无痛、无出血地采样人体皮肤间质液，直接获取人体代谢活跃度最高区域的高分辨率生理信息，并支持在极小尺寸的单芯片上并行检测多种分析物。Yang 强调，该传感芯片已积累了超过 100 万小时的临床验证数据，其读数持续以每 15 分钟一次的静脉采血结果作为基准比对校准。 即时视觉反馈机制与多指标并行监测 (9:52) 在行为干预方面，Yang 分析了连续生物反馈对认知重构的心理学推动作用。以该团队正在进行临床试验的代谢监测设备为例，设备通过位于用户外周视野内的环境光（如 LED 显示屏）提供无感知的实时视觉提示——蓝色表示血糖处于目标范围，黄色则预警血糖偏高。当用户饮用原本以为健康的橙汁后观察到指示灯转黄，便能即刻建立因果关联并触发认知重构。这种低摩擦的即时反馈机制使用户在没有专业教练介入的情况下主动纠偏，例如改喝饮用水或散步 10 分钟以平抑血糖峰值。 在扩展应用方面，该传感平台正从单一葡萄糖监测扩展至更广泛的生理指标。在运动表现领域，传统乳酸阈值监测需要打断运动进行耳垂刺血，而自行车运动员 Sky Christopherson 曾通过连续监测乳酸在不使用违禁药物的前提下打破纪录。Biolinq 在单颗微芯片上集成了多种传感化学体系，成功实现了对葡萄糖与乳酸的同步无交叉干扰检测。在受试者剧烈运动使乳酸达峰并饮用果汁使血糖升高的验证测试中，两类生理信号在微小芯片表面保持了独立与隔离。 合成生物学受体与AI助手的交付闭环 (14:15) Yang 强调，生命科学已步入人工合成生物机制的全新阶段。Biolinq 利用人工合成的 DNA 链工程化构建新型生物受体，开创了此前无法居家连续监测的新应用：其一是用于跟踪苯丙氨酸的连续蛋白质传感器，免除手动记录饮食的繁琐，并辅助使用 GLP-1 减肥药物的患者确保蛋白质充足摄入以防肌少症；其二是连续皮质醇监测，通过跟踪这一压力与认知负荷标志物来优化工作效率并降低系统性炎症；其三是被动防御性贴片，用于在流行病危机中替代鼻拭子，一旦检测到病毒抗原或抗体即可亮灯提示隔离。 针对数据转化为行动的“最后一英里”，Yang 指出仅有海量数据并不足够，必须赋予生物数据生动的人格。现场演示的 AI 健康教练原型“Liz”结合了名人形象与生成式 AI，能够根据穿戴设备的实时读数，以幽默或坚决的口吻督促用户起身运动。这种拟人化交互将复杂的代谢数据转化为对话式干预，旨在有效提高用户的参与度并推动持久的行为改变。 Yang 总结指出，半导体微传感硬件、合成生物受体与生成式 AI 形象的结合，构成了未来个性化健康管理的核心构件，Biolinq 正在圣迭戈推进开放架构平台以推进行业合作。不过演讲中展示的代谢监测设备目前仍处于临床试验阶段，尚未获得 FDA 的正式批准上市；多分析物传感芯片与 AI 虚拟教练在大规模真实世界人群中的实际临床效用，仍有待后续监管审查与市场验证。

## Background 该内容由 Ai4 频道于 2025 年 8 月 27 日发布在 YouTube 平台上。视频简介指出，演讲围绕消费级可穿戴设备数据策略展开，探讨如何结合尖端生物传感器、生成式 AI 化身与医疗大模型，赋予健康数据独特的“个性”，从而激发用户养成更健康的生活方式。 ## TL;DR Biolinq CEO Rich Yang 介绍了全球首款半导体生物传感器芯片，该技术能无痛采集皮肤组织液中的葡萄糖、乳酸、皮质醇及蛋白质等多项高分辨率纵向数据。通过设备灯光直观反馈触发“认知重塑”，并结合具有情感与幽默感的 AI 虚拟教练（如化身“Liz”），打破传统健康数据的枯燥性。这种高精度数据与个性化 AI 的结合，为真正落地精准医疗、促进行为改变及延长健康寿命提供了全新路径。 ## Atmosphere & style 这场演讲气氛热烈且极具感染力，现场互动频繁。Rich Yang 运用实物展示（如晶圆与手臂传感器）和现场突然响起的 AI 虚拟教练干预，打造出跌宕起伏的科技前沿沉浸感。整场分享专业严谨又不失幽默风趣，将硬核深科技突破与心理学行为干预巧妙融合，让听众深刻体会到未来医疗科技充满人性化温度与情感互动的全新魅力。 ## Key takeaways - [00:04](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=4s) 消费级科技与医疗健康的加速融合正在推动行业变革，但高度个性化的精准医疗仍受限于家庭场景高精度纵向数据的匮乏。 - [00:04](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=4s) 现有的医疗 AI 基础模型在影像学之外表现欠佳，核心瓶颈在于无法在日常生活中捕捉高分辨率、高保真度的生理数据。 - [00:43](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=43s) Biolinq 推出了全球首款已提交 FDA 审核的硅基半导体生物传感器芯片，利用纳米级硅材料的坚固特性打造微型传感器阵列。 - [00:43](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=43s) 该技术将半导体工艺与合成分子生物学相结合，通过微侵入皮肤浅层的组织液（ISF）采样，实现无痛、无出血的连续多生物标志物监测。 - [10:16](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=616s) 硬件上的视觉指示灯反馈能在“可教导时刻”触发认知重塑，促使使用者自发做出可持续的健康行为改变。 - [11:23](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=683s) Biolinq 成功在单块芯片上集成了葡萄糖与乳酸双传感器，实现了独立清晰且无交叉污染的连续运动代谢监测。 - [11:23](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=683s) 利用 DNA 序列合成新受体，团队研发出针对苯丙氨酸的连续蛋白质传感器，可有效辅助 GLP-1 药物使用者防止肌肉流失。 - [11:23](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=683s) 通过连续监测皮质醇等生物标志物，可评估人体生理压力与炎症，未来还可扩展至无创病毒抗体检测与防疫报警。 - [17:20](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=1040s) 驱动健康行为改变的最后一公里在于用户参与，将高精度数据与 AI 化身（Avatar）及个性化互动相结合是提升健康寿命的关键。 ## Detailed walkthrough ### [00:04](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=4s) 医疗健康与可穿戴设备的融合及 AI 的局限 演讲者 Rich Yang（Biolinq CEO）以一场互动问答开启了分享，现场询问观众中有多少人在佩戴智能手表、智能手环或连续血糖监测仪（CGM）等生物传感器。他指出，我们正处于消费级科技公司与医疗应用加速融合的创举时代，科技巨头纷纷向 FDA 提交 510(k) 申请。这种融合有望降低医疗成本、改善诊疗效果并开拓全新的增长空间。然而，AI 在医疗健康领域的真正潜力却面临着极大的挑战——实现高度个性化的精准医疗需要来自于家庭场景、持续且高精度的纵向数据（longitudinal data），而这类数据在目前极度匮乏。 目前 AI 在医疗领域表现最亮眼的是医学影像学，例如肿瘤和癌症的检测精度甚至超越了人类医生的肉眼。但在其他健康领域，现有的医疗 AI 基础模型（foundation healthcare AI models）整体表现并不理想。其根本原因在于行业无法在日常生活中捕捉到高分辨率、高保真的身体生理数据，仅凭现有测量指标远不足以支撑起未来的医疗大模型。 - **行业应用范例**：糖尿病管理是利用设备、算法、机器学习和 AI 实现个性化医疗最成熟的领域，其核心可总结为解决两个基本问题：“我目前状况良好吗？”以及“如果不好，我该怎么做？” - **商业价值**：糖尿病领域的生物可穿戴设备（bio-wearables）创造了极高的人均生命周期价值（LTV），不仅在华尔街享有最高的估值乘数，其相关股票更是过去 15 年间标普绩效前 20 的常客。 - **技术瓶颈**：现有的 AI 无法自动发现新的生物标志物。人体中许多至关重要的生物标志物，其浓度往往只有葡萄糖分子的千分之一甚至更低，需要依靠深科技（deep tech）与硬科技（hard tech）的突破。 --- ### [00:43](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=43s) 世界首款半导体生物传感器芯片 针对高精度生理数据获取的难题，Rich Yang 介绍了 Biolinq 研发的全球首款半导体生物传感器（已向 FDA 提交审核）。他展示了一块 8 英寸的晶圆，上面集成有数千个微型传感器。在与现场观众互动讨论钛合金、不锈钢和硅材料的硬度时，他特别强调：在纳米尺度下，硅的硬度实际上超过了钛合金和不锈钢，这使其成为制造超坚固生物传感器的理想基材。 Biolinq 将半导体制造工艺与合成分子生物学相结合，打造出比人类头发丝还要细数百倍的生物受体（bioreceptors），直接贴附于皮肤表面进行连续监测。演讲者以自身参与的 2 型糖尿病临床试验为例，展示了戴在前臂上的可穿戴设备——设备上的蓝色指示灯代表血糖处于目标范围内。如果摄入高糖食物（如提拉米苏或果汁），指示灯会变黄提示血糖飙升。 ``` +-----------------------------------------------------------------+ | Biolinq 半导体生物芯片 | +-----------------------------------------------------------------+ | [硅基芯片] -> [微型电化学阵列] -> [DNA/合成分子生物受体] | +-----------------------------------------------------------------+ | v +-----------------------------------------------------------------+ | 组织液（ISF）微侵入采样 | +-----------------------------------------------------------------+ | - 仅作用于皮肤浅层，不触及神经末梢 (无痛/无血) | | - 实时采集多项代谢生物标志物 (葡萄糖、乳酸、皮质醇、蛋白质等) | +-----------------------------------------------------------------+ | v +-----------------------------------------------------------------+ | 实时反馈与认知重塑 | +-----------------------------------------------------------------+ | - 硬件指示灯 (蓝/黄) -> 视觉直观反馈 | | - 智能手机 App -> 高分辨率纵向数据采集 | +-----------------------------------------------------------------+ ``` - **微型化优势**：芯片尺寸微小到即使放在宝石盒中并配有放大镜也难以用肉眼看清，能够贴合在皮肤浅层且不会触及神经末梢，实现无痛、无出血的无感佩戴。 - **生理学依据**：传感器深入采样的是皮肤组织液（interstitial fluid, ISF）。组织液是人体代谢最活跃的组织之一，包含了几乎所有需要高分辨率监测的生物标志物。 - **临床验证**：该技术已完成超过 100 万小时的临床验证。在试验中，研究人员每隔 15 分钟提取一次患者的静脉血样本，用以与芯片输出的数据进行精准对比。 --- ### [10:16](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=616s) 无感实时反馈与认知重塑 设备带来的实时视觉反馈不仅是数据展示，更是促成行为改变的心理学机制。Rich Yang 指出，当用户饮用一杯橙汁并看到前臂上的传感器由蓝变黄时，这种在“可教导时刻”（teachable moments）发生的无感实时反馈会触发“认知重塑”（cognitive restructuring）。 用户不需要专业的教练或复杂的分析报告，大脑就会自然形成直观联系：“原来喝橙汁会让我的血糖飙升。”这种认知会自发驱动用户做出可持续的健康行为改变，例如改喝白开水，或者在饮用橙汁后散步 10 分钟以平稳血糖。 --- ### [11:23](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=683s) 从最大摄氧量到多标志物并行监测 在运动生理与人体极限性能领域，传统的最大摄氧量（VO2 max）测试过程极其昂贵且繁琐，需要佩戴复杂的呼吸面罩；而在职业足球队中，测量乳酸阈值（lactate threshold）甚至需要中断训练并刺破耳垂采血。线粒体在剧烈运动中负责清除乳酸，一旦超出清除极限就会导致乳酸堆积和肌肉酸痛，但传统手段很难精确捕捉这个动态峰值。 Biolinq 展示了与退役自行车运动员 Sky Christopherson 的合作案例：他凭借数据驱动和连续乳酸阈值传感器，在 35 岁左右的年纪、不使用任何违禁药物的情况下打破了室内场地自行车赛的世界纪录。不仅如此，Biolinq 还实现了在单块芯片上同时集成葡萄糖与乳酸双传感器。在测试中，受试者先进行剧烈运动触发乳酸峰值，随后在乳酸下降时饮用菠萝汁引发血糖飙升，实验证明两个信号独立清晰、无交叉污染（cross-contamination）。 - **合成生物学突破**：通过与顶尖大学合作，利用 DNA 序列合成全新的生物受体，使得以往无法在家庭环境中无创连续监测的物质（如抗体、激素等）变得可测。 - **连续蛋白质传感器**：研发出世界上首款针对苯丙氨酸（phenylalanine）的连续蛋白质传感器。对于使用 GLP-1 类减肥药物的群体，该传感器可精准量化蛋白质摄入量，防止肌肉量流失（sarcopenia）。 - **皮质醇与压力监测**：皮质醇是评估认知功能和心理压力的核心指标。通过监测皮质醇可以评估身体炎症与压力水平，协助打造更智能的健康家居与高效率的工作环境。 - **未来防疫应用**：在面对未来的流行病时，贴片式传感器可替代传统的鼻拭子采样，在接触 COVID-19 等病毒抗原/抗体时直接通过红光报警，提示使用者进行自我隔离。 --- ### [17:20](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=1040s) 终极一步：AI 化身与情感互动 拥有高精度、高保真的大量健康数据只是第一步，如何实现高黏性的“用户参与”（engagement）才是打通健康管理的“最后一公里”。如果只是把海量的数据堆砌给用户，并不能真正促成改变。 在演讲的尾声，Rich Yang 的前臂传感器亮起黄灯，现场突然响起其 AI 健康教练“Liz”的声音：“Rich，我需要你停止讲话，开始走路，你的血糖水平正在升高且持续上涨。”演讲者借此介绍了与领军 Avatar AI 平台合作打造的虚拟教练形象。他总结道，将高精度的生理数据与幽默感、AI 化身（Avatars）以及好莱坞明星 IP 相结合，赋予健康数据独特的“个性”，才是激发人们做出健康决策、延长健康寿命（healthspan）的未来趋势。 <ElicitationsGroup message="如果您想进一步探索 AI 与生物传感技术："> <Elicitation label="深入了解芯片如何在组织液（ISF）中无痛检测多种标志物" query="请详细解释半导体生物芯片如何在组织液中无痛检测多种生物标志物的原理和技术细节。"/> <Elicitation label="探究 AI 虚拟教练如何结合实时生理数据进行行为干预" query="AI 虚拟教练是如何结合连续血糖与皮质醇等实时生理数据来进行心理学行为干预与认知重塑的？"/> </ElicitationsGroup> ## Notable quotes > [17:20](https://www.youtube.com/watch?v=_aWVI3ZO6io&t=1040s) Rich，我需要你停止讲话，开始走路，你的血糖水平正在升高且持续上涨。 — Liz (AI 健康教练) ## Mentioned Rich Yang, Biolinq, Liz, Sky Christopherson, FDA, 标普, 华尔街, GLP-1, COVID-19 ## Video description Rich Yang, CEO, Biolinq Abstracts How data strategies with cutting edge wearables can power AI Health applications to inspire healthier living. Wearable devices are generating large amounts of wellness data, but not enough useful information for consumers. Learn about the future of wearables combined with the power of AI driven avatars and foundation healthcare AI models designed to give health data a personality. With unprecedented high resolution health information captured from home, better engagement and insights are needed to deliver meaningful outcomes and extend human healthspan. [Source](https://www.youtube.com/watch?v=_aWVI3ZO6io)

[17:20] Liz (AI 健康教练): Rich，我需要你停止讲话，开始走路，你的血糖水平正在升高且持续上涨。

[2:25] Rich Yang: Am I okay? [2:25] Rich Yang: What do I do? [9:52] Rich Yang: orange juice made me high [17:20] Liz: All right, Rich, I'm going to need you to stop talking and start walking. [17:20] Rich Yang: Yeah. So, yeah, my sensor is yellow. [17:20] Liz: Your glucose levels are elevated and climbing.

[0:04] Welcome everybody. Glad to be here. It's [0:07] going to be a great intimate session uh [0:09] here and I will be speaking heavily on [0:13] wearables and AI. And so uh to get [0:18] started, how many of you are actually [0:20] wearing a smartwatch right now? How many [0:23] of you keep your hands up? How many of [0:25] you actually use the health information [0:27] from your smartwatch? Okay. How many of [0:30] you are using the smart rings right now? [0:33] Great. And how many of you have actually [0:35] worn a bio sensor before? Continuous [0:40] glucose. Okay. All right. So, you're [0:43] going to get a master's class today on [0:45] bioarables and how AI is going to make [0:48] an impact. So, what what comes next [0:52] after all of the great things that the [0:54] smart watches can provide us? heart rate [0:56] variability, [0:58] pulse occymmetry. Well, what comes next? [1:01] Right? So, for the first time in a long [1:03] time, consumer tech companies are [1:07] converging with medical applications. [1:11] We're seeing consumer tech giants file [1:13] 510ks with the FDA. It's an incredible [1:17] moment to leverage consumer tech and [1:21] advanced applications to lower the cost [1:23] of care and to pro provide better [1:25] outcomes and a platform for growth. And [1:27] AI AI's promise for health care is the [1:31] hardest one. And it's the hardest one [1:34] because hyperpersonalized medicine [1:37] requires highfidelity, high resolution [1:41] information from home continuously with [1:45] longitudinal data and that does not [1:47] exist today. So where AI has been really [1:50] profound in healthcare is for imaging. [1:54] AI could detect tumors and cancers [1:56] better than a trained physician's naked [1:59] eye. But what about everything else? [2:03] Foundation healthcare AI models [2:06] are not good today. And they're not good [2:09] because we can't capture high [2:11] resolution, highfidelity information on [2:13] the human condition from home. So now [2:16] with the things that we can measure, [2:18] they're great, but it's just not good [2:20] enough yet. So this [2:25] is an example of what good looks like. [2:27] So the diabetes domain has been using [2:30] devices, algorithms, machine learning, [2:33] AI for a long time. Personalized [2:37] medicine being delivered for a long [2:38] time. And the lessons learned are [2:41] applicable for everything else we want [2:43] to do in healthcare and big data coming [2:45] from healthcare. So what do I mean by [2:46] this? The best programs for diabetes [2:49] have solutions that help people learn [2:51] about their health, train on better [2:54] behaviors, and track outcomes. [2:58] And tracking in personalized feedback is [3:01] the hardest thing to do in this loop. [3:03] And so, at least for the diabetes space, [3:07] it's a very simple equation. There are [3:10] two things we're trying to answer that [3:12] drives outcomes. [3:16] Am I okay? What do I do? That's it. [3:20] That's all that we need to know. Am I [3:23] okay? And what do I do if I'm not? And [3:27] little do you know or little does [3:31] industry know that at least for [3:33] personalized medicine in diabetes, [3:37] it is the highest lifetime value of a [3:39] customer. [3:41] The bowarables in this domain carry the [3:43] highest valuation multipliers on Wall [3:46] Street. Top 20 performing stock for the [3:49] last 15 years consecutively. [3:51] It is an incredible journey to pair bio [3:55] sensors and bowareables with [3:57] personalized medicine. [4:00] But AI can't detect new biomarkers. At [4:02] least not yet. We're talking about deep [4:05] tech, hard tech. Biomarkers that are [4:09] important to the human condition are a [4:10] thousand times less concentrated than [4:12] the glucose molecule in which for those [4:15] of you that have heard about the glucose [4:16] sensors that people use for general [4:18] health and wellness, diabetes, well [4:20] that's just one marker and it carries [4:22] the highest lifetime value of a customer [4:24] and the highest valuation multipliers in [4:27] healthcare and actually consumer tech. [4:30] Well, this is where we come in. So, we [4:32] are a semiconductor bio sensor company [4:35] and everything I'm about to show you now [4:37] is not yet approved by the FDA, but [4:40] we're excited about where we're going [4:42] with this technology and we have [4:43] submitted to the FDA the world's first [4:45] semiconductor bio sensor. Now, um I have [4:49] two hours worth of content I'm going to [4:50] squeeze into 20 minutes. So, I'm going [4:53] to run quickly and speak really fast, [4:55] but this is the world's smallest bio [4:58] sensor. What you see here is an 8 in [5:00] wafer with thousands of micro sensors [5:04] built into an 8-in wafer. Now, show of [5:06] hands really quickly. Titanium, steel, [5:10] and silicon. These three materials, [5:13] which one is the hardest? Raise your [5:16] hand if you think titanium is the [5:18] hardest material? Raise your hand if you [5:20] think stainless steel is the hardest [5:23] material. Raise your hand if you think [5:24] silicon is the hardest material. All [5:26] right. So, some of you may have heard my [5:28] talk before. Um, so silicon at the nano [5:31] scale is harder than titanium and steel. [5:36] This makes an incredibly robust material [5:39] to build bioensors from. Now, we're [5:42] talking about the deepest hard tech. [5:44] We're in the first era in human history [5:47] where we can now synthesize biology. How [5:49] incredible is that? So, we're leveraging [5:52] semiconductor manufacturing [5:55] sensors built on a silicon chip combined [5:58] with the ability to create new [6:01] biioceptors that are hundreds of times [6:03] smaller than a human hair filament. Put [6:06] it on your skin in measuring biomarkers. [6:08] It is the most incredible hardest deep [6:11] tech you can find. But this is what will [6:14] power potentially [6:16] foundation healthcare AI models. So, [6:19] here is an example of this technology [6:22] being engineered for metabolic health [6:25] for people with type two diabetes. A [6:27] bowarable you wear on the forearm. I'm [6:29] enrolled in a clinical trial here. And [6:31] again, this is not yet FDA approved. And [6:33] it gives me that real-time feedback. [6:35] That blue light, you might not be able [6:36] to see it here, tells me I'm in my [6:38] target range. I had bought a lot of [6:40] junk. Um, I I was going to do this on [6:42] stage, eat some tiramisu, drink some [6:46] orange juice, spike my glucose levels, [6:47] but I'm so full from my dinner last [6:49] night, I just couldn't do it. So, I'm [6:51] blue, which means I'm in my target [6:52] range, and that's good. And so, I get [6:54] that real-time feedback [6:57] on my forearm in my peripheral vision on [6:59] top of higher resolution data on my [7:02] smartphone. [7:06] And there's no audio. So maybe I'll sing [7:09] to this. This is a 60-second clip that [7:13] has no audio. Um, and this is the first [7:16] embodiment of a semiconductor bio [7:18] sensor. Looks like an Apple commercial. [7:20] Um, but so [7:23] this is a breakthrough microal [7:25] technology leveraging. Ah, there we go. [7:37] [Music] [7:54] [Music] [8:00] There it is. Um, I will not and I refuse [8:03] to coordinate a dance to that music. Um [8:06] so but the reason why this is so [8:09] profound is because glucose is just the [8:10] starting point. This platform allows us [8:12] to do the impossible that has not been [8:15] done in the last 30 40 years in [8:17] bowareables. What are the biomarkers [8:19] that you can measure in the skin? So [8:22] there are a lot of biomarkers that are [8:24] available in the skin. And so this was [8:28] why our co-founders of the company that [8:30] came out of UC San Diego PhD program, [8:33] they decided to use silicon chips as the [8:35] start material for this. Why? Because [8:37] it's so robust, it doesn't break. And we [8:39] can make them so small using [8:41] semiconductor manufacturing processes [8:43] that we can sit right on top of the skin [8:44] and we never go deep enough to hit a [8:46] nerve ending. So you can't feel it. No [8:48] bleeding, no pain. And the most [8:50] incredible thing is all of the [8:52] biomarkers that you would want to [8:54] measure at a high resolution [8:56] is found in the interstitial fluid in [8:59] the skin. This is the most metabolically [9:01] active tissue in the body. And this is [9:04] the first time we can actually access [9:06] this compartment of the skin to measure [9:08] bioensors or biomarkers. This wasn't [9:11] possible before because nobody can [9:13] miniaturaturize electrochemical sensors [9:15] this small to access that compartment of [9:17] the of the skin. So, which means we can [9:19] measure multiple analytes in parallel on [9:22] a single chip. And the chip is so small [9:24] you can't even see it. I have a jewel [9:25] case here. There's a little speck right [9:27] in the middle of the jewel case. You [9:29] can't even see it. I have a magnifying [9:30] glass on the other side of it and you [9:32] still can't see it. So, um, in the [9:34] clinical validation here is really [9:36] extraordinary. 1 million hours of [9:39] validation. And what you see here is a [9:40] Venus blood sample in the patient there. [9:42] We're drawing blood every 15 minutes and [9:44] comparing it to the output of our bio [9:46] sensors. the world's first semiconductor [9:48] bio sensor that we actually submitted to [9:50] the FDA and it's under review. So now [9:52] context matters. Um so how do we [9:56] understand our relationship with foods [9:57] we like, exercise, diet, medications we [10:00] take in real time? It's really hard to [10:03] do. And if you haven't worn a bio sensor [10:05] yet, there are versions of this that are [10:08] over the counter now. And metabolic [10:10] healthcare literacy is so important to [10:12] extending human health span. You should [10:14] do it if you haven't tried. So now what [10:16] happens when you have real-time feedback [10:17] passively when teachable moments arise? [10:20] This is why we put this on the forearm [10:21] and you get that real-time feedback. [10:23] Well, let's take a glass of orange [10:26] juice. We all love the vitamin C. We [10:29] love the sweetness. But as I drink that [10:32] orange juice and I see my sensor turned [10:34] from blue to yellow to let me know [10:35] you're high, I got that real-time [10:38] feedback passively when a teachable [10:40] moment happened. And what happens? [10:43] Cognitive restructuring. Cognitive [10:45] restructuring [10:47] is what's required for sustainable [10:49] behavior change. Without a coach, [10:51] without analytics, without data, my [10:53] brain just said orange juice made me [10:56] high. Maybe I can stop drinking orange [10:59] juice. Maybe I could drink water. Maybe [11:01] I could skip orange juice. Maybe I can [11:03] take a 10-minute walk and that will [11:05] bring me back down to my target range. [11:07] But at least I know. So uh and I've got [11:12] few more minutes to run through a lot [11:14] more again two hours worth of data in in [11:18] a master's class on bowarables today in [11:20] AI and I'm going to get to the AI part [11:22] in my last slide. So how do we measure [11:24] other biomarkers then? Well, what is [11:27] this? Anybody know? This is how you get [11:30] V2 max. For those of you that are [11:32] athletes and for those of you that do [11:34] training, this is incredibly expensive [11:37] and difficult and only the most elite [11:39] athletes do this. Now, this is what [11:41] comes out of that. Who can understand [11:44] this, right? I certainly can't. This is [11:46] an easy application for AI, right? But [11:49] this is just not scalable. So, peak [11:51] performance and human performance [11:53] applications for extending human health [11:55] span. It goes on and on and on and on. [11:58] This is the current gold standard for a [12:00] professional soccer team to measure [12:02] their lactate le levels in training. [12:05] They have to stop the practice, stop the [12:07] training, prick your earlobe, draw [12:09] blood, look at the meter, and see where [12:11] your lactate levels are. And why is this [12:12] important? Well, your mitochondria [12:15] clears lactate. If you've ever had a [12:16] hard workout and you were sore, really [12:20] sore the next day, that's lactic acid [12:22] buildup. in your mitochondria. Nobody [12:24] knows when your mitochondria is actually [12:27] at its peak and it can can no longer [12:29] clear lactic acid. Well, if you don't [12:31] know that peak, you can't train to [12:33] improve your threshold. And that relates [12:37] directly to improving your ability to [12:39] improve endurance levels. So, what if we [12:43] can do this? Instead of V2 max and [12:46] everything else, you have personalized [12:49] sports medicine opportunities with [12:50] bowarbs. And this one is a lactate [12:53] sensor that's on the athlete. His name [12:55] is Sky Christopherson. He broke a record [12:58] in the VI drum indoor track cycling [13:01] without doping, without drugs at the [13:04] a ripe young age of what mid30s [13:08] using data and a lactate threshold [13:12] sensor continuously measuring lactate [13:15] levels um being hooked up with [13:17] everything else um alongside of them. So [13:20] now it's pretty incredible the kind of [13:23] data that we're generating off of bio [13:25] sensors. But what if we could combine [13:27] multiple signals on one chip? And we've [13:30] done that. So this has glucose and [13:33] lactate together on a single chip. And [13:36] we're talking about [13:38] these sensors are so small, so closely [13:40] packed together. Well, how is it that [13:43] you put multiple chemistries on [13:45] something so closely together without [13:46] any crosscontamination? So what we do is [13:49] we we ramp up the exercise. So we see [13:53] lactate levels peak and then when they [13:55] start to drop after the exercise, we [13:58] make them drink pineapple juice. And [14:00] then you spike up the glucose levels. [14:02] And you can see those two signals are [14:03] separate from each other and they don't [14:05] peak at the same time, which means [14:07] there's no crosscontamination around a [14:10] very small surface area with a bunch of [14:12] micro sensors next to each other. Well, [14:15] we're in the first era in human history [14:17] to be able to synthesize biology. So, [14:19] we've partnered with leading [14:20] universities to create new bio receptors [14:23] out of DNA. It's actually hard deep tech [14:27] that I mentioned to you earlier. And [14:29] this allows us to measure new analytes [14:32] that were impossible until now. Why [14:35] don't we see COVID antigen antibody [14:39] sensors? Why don't we see hormone [14:42] sensors? Well, because there is no [14:44] technology that can measure this from [14:47] home passively continuously without [14:50] blood and without pain. [14:54] But for the first time, we can access [14:55] the skin and do this. So what you see [14:58] here for phenyl alanine, very technical [15:00] name, but that is the world's first [15:02] continuous protein sensor. Instead of [15:04] taking food logs and pictures of the [15:06] foods that you're eating, what if we had [15:08] a sensor that can actually quantify the [15:11] amount of protein that you're ingesting? [15:13] And why does that matter? Sarcopenia, [15:14] for those of you that may know, for [15:16] people that are on GLP1s [15:18] with lean muscle mass loss, that's not [15:21] good. If you know how much protein [15:23] you're taking or not, you can provide [15:25] some recommendations for how much [15:26] protein you should be eating to mitigate [15:28] lean muscle mass loss and still have the [15:30] benefits of the weight loss from the [15:32] GLP1s. It's incredibly important. If I [15:35] don't have to manually log foods and I [15:37] know the glycemic index of what you're [15:39] eating, I know the amount of protein [15:41] that you're ingesting, what next? [15:45] And this is the power of AI that will be [15:47] coming. And so now cortisol this is the [15:51] single most important biioarker that [15:54] everybody is asking us to pursue. Why [15:58] cortisol is a surrogate marker for [16:01] cognitive function and stress. [16:05] And so I would argue we put everybody in [16:08] our company on the sensors their [16:09] cortisol levels are probably skyhigh. [16:11] Right? So maybe not good for me but in [16:13] our team but cortisol is a biioarker now [16:16] that is actually possible to measure [16:18] using this technology. And so if we can [16:21] mitigate inflammation and reduce stress [16:24] how great would this world be? And I'll [16:26] tell you, it would be great if I knew [16:27] that my employees were stressed. Then I [16:30] buy everybody happy hour for the day, [16:31] right? So, um, more productive work [16:34] force, healthier workforce, but this [16:36] allows us to make a smart home, a [16:40] smarter healthcare environment at a much [16:42] lower cost. Um, and again, you have to [16:46] understand your relationship to foods, [16:48] medicine, activity, sleep, and [16:51] preferably all in one wearable device to [16:54] really help extend human health span, [16:56] right? But what about the next pandemic? [16:58] Instead of nasal swabs, bowarables will [17:02] be on the front line. What if I can get [17:04] a red light that shines if I detect and [17:08] sense that I've been exposed to a COVID [17:10] antibbody and antigen and I can have [17:11] this patch on my body for two weeks and [17:14] I see a red light means I'm exposed. I [17:16] can now go self- quarantine, go through [17:18] the therapeutics, that would be [17:20] extraordinary. But the final mile is the [17:23] most difficult one. These are all great [17:24] technologies, a lot of data, high [17:26] resolution, highfidelity data. But [17:30] engagement is the final mile. And what I [17:34] mean by engagement. [17:35] >> All right, Rich, I'm going to need you [17:37] to stop talking and start walking. [17:39] >> Yeah. So, yeah, my sensor is yellow. [17:41] >> Your glucose levels are elevated and [17:43] climbing. [17:45] >> So, that's Liz, my AI wellness coach. [17:49] That's her avatar. Liz is actually right [17:50] here sitting right here with me and [17:52] she's part of a group of companies that [17:55] are on the leading edge of Avatar AI um [17:59] coaching platforms but they have [18:02] something really unique A-list [18:04] celebrities behind them. So stay tuned [18:06] and see see more uh for that. And again [18:09] I've got one minute left and what I [18:11] wanted to say is the engagement is [18:14] what's going to drive outcomes. We have [18:15] too much data that we're generating not [18:18] enough information. how we present that [18:20] information using humor, AI avatars, [18:23] coaching, celebrities, even that would [18:27] inspire me to make healthier decisions. [18:31] This is the call to action from groups [18:34] here. Come join our mission. We're based [18:36] in San Diego and we're going to have an [18:38] open architecture biosensing platform to [18:41] do the impossible and extend human [18:43] health span. So, thank you for your