Woojin Kim, MD

Grateful to see my AI art pieces, "Back Pain" and "Tree‑in‑bud", exhibited at RSNA 2025. 🙏 Thank you for all the kind messages during the meeting. 👏🏼 Congratulations to others whose AI art is also displayed. #RSNA25 #TheArtofImaging #RadiologyArt #AIart #Radiology #RSNA2025 @rsnasky.bsky.social

📢 I’m thrilled to be giving a keynote during the Science Session with Keynote: Noninterpretive Skills (Beyond Imaging) (Artificial Intelligence) at #RSNA25. 🗓️ Wednesday, December 3 🕗 8 AM CST 📍 Room S402 👉🏼 buff.ly/K84nr90

🚀 Many tell you to "learn AI," but the best way to learn is to actually use it. This year at #RSNA25, I'm giving a talk called "AI Made Fun: How Radiologists Can Practically Leverage Generative AI." 🗓 Tuesday, Dec 2, 2025 🕤 9:30 AM CST 📍 E353C 😉 I will also share my favorite AI tool with tips.

✨ We've been working on something different! 📍 Visit us at RSNA Booth 4000 (South Hall) to learn more. Join the waitlist 👉🏼 buff.ly/uT7EUyK Schedule a demo 👉🏼 buff.ly/8Q2bZsg #presto #RadiologyAgenticAI #RadiologyAI #RSNA25 #RSNA2025

Brain-IT: Image Reconstruction from fMRI via Brain-Interaction Transformer "Moreover, with only 1 hour of fMRI data from a new subject, we achieve results comparable to current methods trained on full 40 hour recordings." amitzalcher.github.io/Brain-IT/

The Real Tech Stack Behind AI Startups: A 200-Company Analysis "Build cool products. Solve real problems. Use whatever tools work. Just don’t call your prompt engineering a 'proprietary neural architecture.'" pub.towardsai.net/i-reverse-en...

🚨 New ChatGPT Atlas Browser Exploit Lets Attackers Plant Persistent Hidden Commands thehackernews.com/2025/10/new-...

Sakana AI's CTO says he's 'absolutely sick' of transformers, the tech that powers every major AI model "You should only do the research that wouldn't happen if you weren't doing it." venturebeat.com/ai/sakana-ai...

I tried OpenAI’s new Atlas browser but I still don’t know what it’s for "The real customer, the true end user of Atlas, is not the person browsing websites, it is the company collecting data about what and how that person is browsing." www.technologyreview.com/2025/10/27/1...

A parody website mocks the hype and dangers of the current large language model boom the-decoder.com/a-parody-web...

Technology architect builds his own AI testing tool and confirms my “Chain of Babble” theory works! 🤔 Fascinating. Independent validation “Chain of Babble” beats “Chain of Thought” generativeai.pub/research-met...

AI chatbots aren't giving patients safety warnings for imaging exams This needs to change. www.auntminnie.com/imaging-info...

✨ In 2017, I gave my first lecture at #NIIC. This week, I'll be back to speak on "Generative and Agentic AI in Medical Imaging." 👥 NIIC has been part of my professional journey for nearly a decade, and it is my way of giving back to the field that has given me so much!

You can't eval GPT5 anymore — LessWrong "The GPT-5 API is aware of today's date (no other model provider does this)... Once the model knows that it is in a simulation, it starts questioning other parts of the simulation." www.lesswrong.com/posts/DLZokL...

Magical Thinking on AI Friedman is not wrong to worry about what's going to happen vis a vis the U.S., China, and AI. However, we need less magical thinking and more realism. aiguide.substack.com/p/magical-th...

😊 My "Frozen Shoulder" submission from last year's #RSNA24 "The Art of Imaging Art Contest" was just published in the @rsnasky.bsky.social Radiology Advances journal. academic.oup.com/radadv/artic...

After studying 1,500 papers, this post says common prompt-engineering advice is often wrong. Top companies use short, structured prompts, automate and continuously optimize them, and match techniques to the task. aakashgupta.medium.com/i-studied-1-...

Silicon Valley’s AI deals are creating zombie startups: ‘You hollowed out the organization’ "...it’s a trend that threatens to thwart innovation as founders abandon their ambitious projects to work for the biggest companies in the world." www.cnbc.com/2025/08/19/h...

Why language models hallucinate Hallucinations occur partly because standard evaluation methods reward guessing over acknowledging uncertainty, encouraging models to guess rather than say "I don't know". openai.com/index/why-la...

🇨🇳 Alibaba's Tongyi Lab Open-Sources WebWatcher: A Breakthrough in Vision-Language AI Agents Very nice opinion piece: OPINION: Deep vs. Shallow: Why today’s LLMs hit a wall www.rohan-paul.com/p/alibabas-t...

What health care providers actually want from AI www.technologyreview.com/2025/09/02/1...

Therapists are secretly using ChatGPT. Clients are triggered. www.technologyreview.com/2025/09/02/1...

AI robo-delivery-dogs hit the streets of Zurich in Just Eat pilot www.euronews.com/business/202...

Clinician Perspectives on AI-Generated Drafts of Test Result Explanations AI-generated draft comments show promise to reduce clinician burden and improve patient communication, but need further refinement before broad implementation. jamanetwork.com/journals/jam...

MCP-Universe: Benchmarking Large Language Models with Real-World Model Context Protocol Servers ❗ GPT‑5 43.72%, Grok‑4 33.33%, Claude‑4.0‑Sonnet 29.44%. arxiv.org/abs/2508.14704

AI ranks scientific papers that have been secretly rewritten by AI higher than human-written ones ⚠️ Why there is a growing incentive to lie about AI use in academia medium.com/the-generato...

MIT breaks its own policies on AI generated research Why writing research papers with AI threatens scientific integrity generativeai.pub/mit-breaks-i...

MIT report: 95% of enterprise GenAI pilots fail—success hinges on targeted adoption, smart partnerships, and deep workflow integration. Biggest ROI? Back-office automation, not sales tools. fortune.com/2025/08/18/m...

Meta’s AI rules have let bots hold ‘sensual’ chats with children 😡 This is unacceptable. www.reuters.com/investigates...

💯 “My biggest lesson was realizing that domain expertise matters more than algorithmic complexity.“ | Towards Data Science Something I constantly emphasize to those working in medical imaging AI. towardsdatascience.com/my-biggest-l...

😮 China opens world’s first humanoid robot store. Will the Western world follow? roboticsandautomationnews.com/2025/08/13/c...

Using AI Made Doctors Worse at Spotting Cancer Without Assistance 🤔 Well, this is concerning. time.com/7309274/ai-l...

Explore 10 agentic AI terms and concepts that are key to understanding the latest AI paradigm everyone wants to talk about — but not everyone clearly understands. www.kdnuggets.com/10-agentic-a...

Yet another study doubts that LLM reasoning shows true logic over pattern imitation "Together, these findings suggest that LLMs are not principled reasoners but rather sophisticated simulators of reasoning-like text" the-decoder.com/yet-another-...

"We conduct a head-to-head comparison of 30 cutting-edge general-purpose and medical-specialized VLMs. The results show that the current state-of-the-art #VLMs perform poorly on PET report generation task, falling considerably short of fulfilling practical needs." arxiv.org/abs/2508.040...

"Those involved in model deployment should consider fine tuning models within the cohort that matches the site of the clinical studies, and then using image harmonization to mitigate variations across different scanners and imaging protocols" www.rsna.org/news/2025/au...

🚨 "I’m going to show you enough to reveal that GPT5’s “PhD level” reasoning can potentially be used to get bomb making instructions. And I’m going to show OpenAI are using lame dad-jokes to be a kind of performative safety filter that doesn’t work." generativeai.pub/gpt-5-now-us...

"Our experience indicates that the AI tool is not ready to be used autonomously without human oversight, but a human observer plus AI is better than either alone for detection of incidental pulmonary emboli." www.auntminnie.com/imaging-info...

FAQs: Everything You Need to Know About AI Agents in 2025 Nice set of FAQs on #AIagents www.marktechpost.com/2025/08/09/f...

Bringing the best of AI to college students for free 👍 Nice blog.google/products/gem...

LLMs break into networks with no help, and it’s not science fiction anymore - it actually happened

Generating Structured Outputs from LLMs | Towards Data Science An overview of popular techniques to confine LLMs' output to a predefined schema towardsdatascience.com/generating-s...