On this episode of Advances in Care , host Erin Welsh and Dr. Craig Smith, Chair of the Department of Surgery and Surgeon-in-Chief at NewYork-Presbyterian and Columbia discuss the highlights of Dr. Smith’s 40+ year career as a cardiac surgeon and how the culture of Columbia has been a catalyst for innovation in cardiac care. Dr. Smith describes the excitement of helping to pioneer the institution’s heart transplant program in the 1980s, when it was just one of only three hospitals in the country practicing heart transplantation. Dr. Smith also explains how a unique collaboration with Columbia’s cardiology team led to the first of several groundbreaking trials, called PARTNER (Placement of AoRTic TraNscatheteR Valve), which paved the way for a monumental treatment for aortic stenosis — the most common heart valve disease that is lethal if left untreated. During the trial, Dr. Smith worked closely with Dr. Martin B. Leon, Professor of Medicine at Columbia University Irving Medical Center and Chief Innovation Officer and the Director of the Cardiovascular Data Science Center for the Division of Cardiology. Their findings elevated TAVR, or transcatheter aortic valve replacement, to eventually become the gold-standard for aortic stenosis patients at all levels of illness severity and surgical risk. Today, an experienced team of specialists at Columbia treat TAVR patients with a combination of advancements including advanced replacement valve materials, three-dimensional and ECG imaging, and a personalized approach to cardiac care. Finally, Dr. Smith shares his thoughts on new frontiers of cardiac surgery, like the challenge of repairing the mitral and tricuspid valves, and the promising application of robotic surgery for complex, high-risk operations. He reflects on life after he retires from operating, and shares his observations of how NewYork-Presbyterian and Columbia have evolved in the decades since he began his residency. For more information visit nyp.org/Advances…
背景阅读 神经搜索 听起来很抽象,其实典型的神经搜索系统就是大家在电商平台上常用的 以图搜图 , 智能问答机器人 等等。神经搜索可以用于提高搜索视频、图片和音频等等非结构化多模态数据的效率。例如,可以使用神经搜索来搜索家庭影集中的图片或视频。使用神经搜索,用户可以通过语音、图片或描述文字来搜索文件,而不必按照传统的方式逐个查看文件名。这可以大大提高搜索效率,使用户能够更快地找到所需的文件。 但是搭建这样一套系统是非常耗时耗力的,而 Jina AI 极大简化了基础设施的复杂性。不管是想做一套基于神经搜索的完整解决方案,还是想在现有解决方案上加入神经搜索功能,当开发者和企业需要时,Jina AI 已经将从构建到部署所有的技术栈都准备好了。 嘉宾 王楠 ,Jina AI 联合创始人兼 CTO,博士毕业于德国波鸿鲁尔大学。 自 2009 年开始从事深度学习相关研究,之后先后担任德国知名电商 Zalando 高级数据科学家,腾讯高级研究员,在搜索和推荐领域的具有丰富的模型设计、实现和部署经验。 主持人 Mikey 开源爱好者、一个码字的 Rick 开源爱好者、业余开源布道者 时间线 01:24 嘉宾王楠自我介绍 02:32 王楠回顾生成式 AI 崛起的历程 08:38 模型到底是什么?它的本质是一个函数。 12:30 将文字转化成向量,背后原理是什么 18:00 对于不懂 AI 原理的开发者,有哪些开箱即用的 AI 工具 20:20 Jina AI 创立的初衷:极大简化处理 AI 基础设施的复杂性,帮助开发者和企业快速将 AI 模型落地。 26:53 AI 已经过了从 0 到 1 的时代,现在是从 1 到 N 的时代。 27:31 CTO 解读什么是神经搜索,和传统搜索有什么区别 33:40 王楠介绍神经搜索技术的落地普及度 36:20 神经搜索的落地场景:通过语音、图片或描述文字来搜索家庭影集的图片和视频 50:08 产品最新动态:正式将 DocArray 项目捐赠给 Linux 基金会 62:02 未来的 AI 是能够跨越领域和行业的界限,并被应用于各种不同的场景。 69:50 作为 CTO,最希望团队能一直保持好奇心。Staff Rick 开源爱好者、业余开源布道者 扩展阅读 DocArray ,非结构化数据的数据结构 Jina ,搭建多模态、跨模态应用的云原生 MLOps 框架 Jina AI 创始人肖涵博士解读 多模态 AI 的范式变革…