最近のAI関連研究で気になったものをピックアップ〜2025年12月の生物医学論文の約9割にLLM支援の痕跡が見られるという分析が出た。arXivプレプリントとNature報道で確認。特徴的なのはセクションによる差だ。Discussion部分で痕跡が高く、Methodsは相対的に低め。「考察を書く」部分ほどLLMに依存し、「実験方法」はまだ人間が書いている割合...
🤖 Made a tool to remove SynthIDs from imagesAs you know, whenever you edit an image via Gemini or OpenAI, they plaster a SynthID to mark it as their own. Further, these SynthIDs ca...
Claude Academy is a great new resource for anyone looking to become more effective with Claude AI.It offers courses, tutorials, and practical use cases covering everything from Cla...
ユグドラシルでもmodelのことは話題になっているかなThe oldest iPhone model you can trade-in with Apple is older than you'd think https://www.engadget.com/2240134/oldest-iphone-model-trade-in-with-apple/#A...
🐝 #SwarmForge is an #opensource agent coordination system that turns swarms of #AI agents into a disciplined engineering team, built on #tmux, #git worktrees and #Babashka #devtool...
I gave my team full context this quarter.Org charts. Decision logs. The Q3 2023 binder.They stopped making decisions.They had too much information to act on any of it.I revoked the...
Can a startup called Outer Biosciences figure out how to keep living human tissue alive outside the body? Read more from @Techcrunch:https://flip.it/F-ZN2H#Tech #AI #Technology
Google just bought 40 years of Spirit Airlines' internal data for $10M to train AI. Customer records were excluded — but 175,000 employee files were not. A flight attendants union ...
Using AI Slop to Sell Cameras Will Never Make Sense to Mehttps://petapixel.com/2026/08/22/using-ai-slop-to-sell-camera-equipment-will-never-make-sense-to-me/#Photography #AI #Opini...
I wonder how well this #captcha works these days.https://github.com/rauchg/doom-captcha#scrapers #LLM
@cwebber:“The term #GenAI points at exactly what it's worst at: generating things. If we want to talk about finding problems, it's a different story. But even leaving aside the qua...
Today I'm furious. Heavily furious. No, it's not the victim tone but the angry tone. I escaped from Facebook and X because of far-right, and on Fediverse I have to argue with far-l...
Example of #AI "artificial stupidity"During the 2025 war between Iran and Isra-hell, someone in Iran shared an image that was supposed to be a shot-down IOF F-35 but showed an impo...
Ah, the thrilling spectacle of yet another #AI "speedrun" where they pretend to break records no one else has heard of! 🚀💨 153 runs and 18 #models later, we're still left pondering...
NanoGPT Speedrun Frontierhttps://www.primeintellect.ai/research/nanogpt-speedrunComments: https://news.ycombinator.com/item?id=49404380#HackerNews #NanoGPT #Speedrun #Frontier #AI ...
🚀 Fastest-growing AI projects today1. The focus on FPGA-based inference chips and deep integration with robotics-first visual...2. Additionally, community-driven projects around do...
📰 Fighter jets help destroy Russian drone boat near European offshore gas platformKaynak: Ars Technicahttps://arstechnica.com/gadgets/2026/08/explosive-russian-drone-boat-destroyed...
2026-08-21 | 💑 Gravitational Collapse 💑#AI Q: 🌌 When does seeking space become an act of abandonment?🧠 Attachment Styles | 🧱 Communication Barriers | 🌩️ Relationship Conflicthttps:...
Apple Cuts 200+ Jobs Across Vision Pro and Siri TeamsApple lays off more than 200 employees across its Vision Pro, Siri, and immersive video teams, largely shutting down its Vision...
NanoGPT Speedrun Frontierhttps://www.primeintellect.ai/research/nanogpt-speedrun#ai
Should AI Agents be ACID compliant?AI agents are becoming increasingly autonomous, but autonomy introduces a fundamental systems problem: how do we make sure an agent’s actions rem...
#meme #bias #information #ai #technology #feed #media #future #software #problem
The #TCSAI—THE DAWN OF A NEW CIVILIZATION: Sustainable, Ethical and Harmonized Scientific Capitalism (SCHEC). This is not just a book. It is a civilizational manifesto, a practical...
Anthropic is deploying its sensitive Mythos 5 model for enterprise security, but there is a catch: customers never get direct prompt access. By delivering only vulnerability scans ...