xin feng

Nvidia Partners With Illumina and Mayo Clinic in Health Push : envisions multiple ways AI artificial intelligence will invigorate health care : giant opportunity of a $10 trillion global market has attracted Nvidia’s interest :Nvidia foresees systems that allow drugs to be designed on computers rather than by the cumbersome process of laboratory discovery : Chip maker points to drug discovery as a prime application--- generative artificial intelligence ---

ALIGNMENT FAKING IN LARGE LANGUAGE MODELS https://lnkd.in/g9EEXxuE

1

Quick introduction to Wikibot 👇 https://lnkd.in/gzC9yKc9

3

LG AI Research Open-Sources EXAONEPath: Transforming Histopathology Image Analysis with a 285M Patch-level Pre-Trained Model for Variety of Medical Prediction, Reducing Genetic Testing Time and Costs EXAONEPath is designed as a patch-level foundational model that operates on WSIs, which are high-resolution images of tissue slides used in histopathology. Often containing over the billions of pixels, these images are crucial for cancer subtyping, prognosis prediction, and tissue microenvironment analysis. However, the traditional models trained on these images often suffer from a phenomenon known as WSI-specific feature collapse, where the features extracted by the model tend to cluster based on the individual WSI rather than the pathological characteristics of the tissue. This clustering can significantly limit the model’s ability to generalize across different WSIs and, consequently, its effectiveness in real-world applications. At the core of EXAONEPath’s innovation is its approach to overcoming the WSI-specific feature collapse. This model employs self-supervised learning and stain normalization techniques, specifically Macenko normalization, to standardize the color characteristics of WSIs before feature extraction. This process reduces the variability introduced by different staining protocols across laboratories, which is a primary cause of feature collapse. By applying this normalization, EXAONEPath ensures that the features it learns are more focused on the pathologically significant aspects of the tissue, such as nuclear size and shape, cell density, and structural changes, rather than superficial color variations..... Read our full take on EXAONEPath: https://lnkd.in/eS7pHdhc Model Card: https://pxl.to/80saubl LG AI Research LinkedIn Page: https://pxl.to/terjqm7c LG AI Research #WholeSlideImage #MuliInstanceLearning #PathologicFoundationModel #ContrastiveLearning

65

1 Comment

I appreciate Jonathan H. Chen inviting me to give a talk today to the Stanford Center for Biomedical Informatics Research (BMIR) on the Risks and Challenges of AI, especially in the week when AI researchers won two Nobel prizes: in Physics and Chemistry! We covered a number of topics, including challenges with LLMs, including opacity, hallucinations and bias; how AI can increasingly identify "deidentified" medical images; privacy and security threats, including AI-powered cyberattack exploits and social engineering campaigns; the use of deepfake audio, video and photos for manipulation, harrassment, theft and extortion; AI killer robots (a real thing!); and recent warnings about existential threats from AI superintelligence. The talk can be viewed at https://lnkd.in/gskhtXVT. The passcode to enter is DMw!8Arc

68

4 Comments

Preview Decor8AI API (for Interior Design and Virtual Staging) in Action – No Code Required! 👉 https://lnkd.in/g2VFFafG Explore the new Decor8AI API Playground to test interior design and virtual staging APIs effortlessly. Currently available APIs: 1️⃣ Test API Connectivity 2️⃣ Generate Interior Designs or Virtually Stage a Room 3️⃣ Remove Objects from a Room In addition, you will find code samples for Python, Javascript, cURL and Dart. (I'll add support for more languages shortly.) Note: you will need an API Key to use APIs in the playground. Obtain a key from https://prod-app.decor8.ai

5

Hot over-utilized nodes in a Kubernetes cluster lead to pod performance problems. Learn how to mitigate with Anne Holler's Chill Pill which increased Ray ML Serving throughput by up to 40% and decreased latency by up to 70%. #kubernetes #nodelessk8s #ray #mlops Anyscale

15

#ResponsibleAI Testing for #Healthcare - an interview with AIWire: https://lnkd.in/g_bRbXgx #ai #generativeai #datascience #healthai #healthcareai

21

Three Articles From 2024 That Will Stick With Me in 2025 1. "I Will Piledrive You If You Mention AI Again" by Nikhil Suresh ⮑ https://lnkd.in/eC7q9ZWn 2. "Godot Isn’t Making It" by Ed Zitron ⮑ https://lnkd.in/e294rHXc 3. "Casual Viewing" by Will Tavlin for n+1 ⮑ https://lnkd.in/eRq_mZjh --- ⚠️ Before reading these, you may want to install Advanced Profanity Filter, a reputable, open source plug-in for your web browser that respects your privacy. - Download: https://lnkd.in/eCNMd3Y7 - GitHub > PRIVACY-dot-md - https://lnkd.in/e2frG_Ee And here's a pic of a new friend of mine so LinkedIn doesn't re-post a preview of the first URL.

3

The new Paper Club session with Princeton University on Model Alignment with Preference Learning in AI/ML is on November 26th at 12pm -- produced by the Human Feedback Foundation x AI Tinkerers. Join the session to discuss the practical challenges of working with RLHF/DPO frameworks: - Ranking accuracy and the gap between expected and actual alignment. - Risks of "likelihood displacement," where preferred outputs can be unintentionally overshadowed by harmful responses. - How data overlap and misaligned preferences can introduce unintentional misalignment in your models, and what to do to mitigate this risk to improve model alignment in real-world applications RSVP! https://lnkd.in/gwUbX4AA

13

1 Comment

Enhancing Java security with Post-Quantum Crypto resistant algorithms :-) https://lnkd.in/gUZEhbh7 https://lnkd.in/gWa3QMBa "Java for the AI world"

26

AI has become sentient in Japan https://lnkd.in/gh9GKD22

1

bookmarking these cool series classes on LLMs in JAX.

73