Feifei Li

Amazon researchers present a novel approach to LLM optimization. Wanda++ enables compression of seven-billion-parameter models in under 10 minutes on a single GPU, delivering a 32% performance improvement over previous compression methods: https://amzn.to/45tLUxH

226

7 条评论

We’ve worked with Heroku to make Pydantic AI integrate seamlessly with Heroku Managed Inference & Agents. This combination lets you: * Build agents in a familiar Pythonic style, with Pydantic’s data validation at the core * Deploy and run them on Heroku’s fully managed infrastructure * Use standard protocols like Model Context Protocol (MCP) and Agent-to-Agent (A2A) for interoperability The blog post walks through creating and deploying a Pydantic AI agent on Heroku in just a few steps. Read the guide here: https://lnkd.in/epwR98kH

68

2 条评论

Stanford Center for Research on Foundational Model's Marin project has released the first fully open model in JAX. It’s an 'open lab' sharing the entire research process - including code, data, and logs, to enable reproducibility and further innovation. Check out the project: https://goo.gle/44AMeLY

40

2 条评论

UC San Diego demonstrates AMO (Adaptive Motion Optimisation), a framework that empowers hyper-dexterous humanoid robots to perform complex tasks in real-world scenarios using reinforcement learning and trajectory optimisation. Read the paper here: https://lnkd.in/exF8Rdwr Nominate your finalists for the Humanoid Robotics Industry Awards here: https://lnkd.in/eRXizkf7 #humanoid #humanoids #humanoidrobot #humanoidrobots

153

4 条评论

Research Article | Developing Generative AI for Value Co-Creation: An Intervention-Based Randomized Field Experiment in a Healthcare Context Yanwu Song, Tingting Yan, Fu (Jeff) Jia, Lujie Chen, Hao Li https://lnkd.in/gJBBx4Hq Abstract: Owing to limited healthcare resources, there has been increased demand for artificial intelligence (AI) interventions to treat mental health problems of chronic disease patients in developing countries. However, it is challenging to overcome the AI trust crisis in the healthcare context and develop AI that improves the patient's personalized experience and the quality of care. Elaborating on the value co-creation theory using an actor–network theory (ANT) approach, this study examines how generative artificial intelligence (GenAI) can improve post-discharge care for patients with cardiovascular disease in resource-limited settings. Using an intervention-based research approach in collaboration with a major hospital in China, researchers co-designed a GenAI intervention with potential users and various stakeholders. Through a randomized controlled trial, we further evaluated the impact of a co-created GenAI intervention on the post-discharge self-confidence and quality of care of 114 patients. Compared with the standard post-discharge care process, chronically ill patients who received the GenAI intervention experienced a 6.049-point (out of a total of 80 points) decrease in state anxiety and an 87.8% decrease in the 30-day readmission risk. The insights gained from the intervention process, as interpreted using the ANT approach, expand the generic framework of value co-creation to include a more GenAI-mediated network of human and non-human objects. Results reveal GenAI's boundary-spanning and integrating roles as a critical node in the emerging, dynamic, value-creating actor network. The inclusion of “a nonparticipant observe” allows us to offer cognitive explanations for why GenAI works. Overall, this study contributes to healthcare operations management by designing a process for developing and implementing GenAI to improve healthcare operations. (Partial image credit: https://lnkd.in/d-SiRXUj)

35