“Can AI meaningfully help with bioweapons creation? On our new Virology Capabilities Test (VCT), frontier LLMs display the expert-level tacit knowledge needed to troubleshoot wet lab protocols. OpenAI’s o3 now outperforms 94% of expert virologists. https://x.com/DanHendrycks/status/1914696657813561799
Johnson & Johnson Pivots Its AI Strategy – WSJ https://archive.md/EzUQD
“Turns out, LLMs represent numbers on a helix and use trigonometry to do addition. A new paper reverse engineers addition in models like GPT-J-6B and finds a “Clock” algorithm. Numbers are encoded using sine and cosine terms, then added like angles. https://x.com/LiorOnAI/status/1914334179929530660
“I’m hearing more and more stories of ChatGPT helping people fix longstanding health issues. We still have a long way to go, but shows how AI is already improving people’s lives in meaningful ways. https://x.com/gdb/status/1914106403574452496
“We built an AI model to simulate how a fruit fly walks, flies and behaves – in partnership with @HHMIJanelia. 🪰 Our computerized insect replicates realistic motion, and can even use its eyes to control its actions. Here’s how we developed it – and what it means for science. 🧵 https://x.com/GoogleDeepMind/status/1915077085325922785
AI assisted search-based research actually works now https://simonwillison.net/2025/Apr/21/ai-assisted-search/
“For the first time, I truly believe AI might replace doctors. I felt dizzy every morning when standing up from bed, and even worse after moving to a hill in SF, the dizziness hit me every time I climbed. Visited my primary care doctor twice; both times, they said it’s normal,” / X https://x.com/Yuchenj_UW/status/1914000352606818419
“First we map the weather. Then we predict it. Eventually — we control it. Same curve for every phenomenon. So who decides tomorrow’s forecast? https://x.com/bilawalsidhu/status/1913416236752408870
“deploying fusion to the world is surprisingly gated on high-power semiconductors. the most important place you could design and build those high-power semiconductors (and do other awesome electrical engineering) is helion. please consider: https://x.com/sama/status/1913609769518317929
“OmniV-Med: Scaling Medical Vision-Language Model for Universal Visual Understanding “First, we construct OmniV-Med-Instruct, a comprehensive multimodal medical dataset containing 252K instructional samples spanning 14 medical image modalities and 11 clinical tasks. Second, we https://x.com/iScienceLuvr/status/1914624649142657199
“What could scaling unlock for biology? Introducing ProGen3- our next AI foundation models for protein generation. We develop compute-optimal scaling laws up to 46B parameters on 1.5T tokens with real evidence in the wet lab. +we solve a new set of challenges for drug discovery https://x.com/thisismadani/status/1912502918328315937
[2502.03544] Gold-medalist Performance in Solving Olympiad Geometry with AlphaGeometry2 https://arxiv.org/abs/2502.03544
TerraMind: Large-Scale Generative Multimodality for Earth Observation https://arxiv.org/pdf/2504.11171
[2410.04983v1] RoWeeder: Unsupervised Weed Mapping through Crop-Row Detection https://arxiv.org/abs/2410.04983v1
“How new data permeates LLM knowledge and how to dilute it Author’s Explanation: https://t.co/pxiCjPTpVR Overview: This research investigates how new information integrates into LLMs, identifying a “priming” effect where learning specific facts leads to their inappropriate https://x.com/TheAITimeline/status/1914175959751684400
“Biotech co Profluent dropped ProGen3 models for complex protein design —Flagship 46B LLM trained on 3.4B+ protein sequences —Designed antibodies matching approved therapeutics — Also created gene editing proteins half the size of CRISPR-Cas9 https://x.com/rowancheung/status/1914201334519124080
“NEMOTRON-CROSSTHINK: Scaling Self-Learning beyond Math Reasoning “In this work, we propose NEMOTRON-CROSSTHINK, a framework that systematically incorporates multi-domain corpora, including both synthetic and real-world question-answer pairs, into RL training to improve https://x.com/iScienceLuvr/status/1914595485148701079
“so you guys know RL can be used for more than just math and coding, right?” / X https://x.com/iScienceLuvr/status/1913783561725100494
“As AI models become more complex and more capable, is it possible that they’ll have experiences of their own? It’s an open question. We recently started a research program to investigate it. https://x.com/AnthropicAI/status/1915420604397744497
“Introducing Open Deep Research in TypeScript! This is a rewrite of our Python implementation to TypeScript specifically made for web devs. Try the script today. We also have a CLI and full-stack OSS deep research app that are coming soon! https://x.com/togethercompute/status/1914721242285838498
“New research finds models are “only a few tasks away” from autonomously replicating (spreading copies of themselves without human help) RepliBench, a benchmark that measures the abilities of frontier AI systems to autonomously replicate, shows it’s not “if” LLMs self-replicate, https://x.com/rohanpaul_ai/status/1915145069188853778
“More Research on integrating tool use in long-form reasoning! ReTool teachs LLM to dynamically interleave reasoning with tool-use (code execution) using Reinforcement Learning. Implementation: 1️⃣ Create math reasoning data and automatically convert calculation steps into code https://x.com/_philschmid/status/1913869660463702435
“[ICLR] AI research hot takes Day 1: It’s time to go from R to D. https://x.com/shaneguML/status/1915169621042499846
“Static scientific diagrams often lack editable structure, hindering reuse. This paper introduces “Draw with Thought” (DwT), a framework guiding Multimodal LLMs (MLLMs) to reconstruct diagrams into editable mxGraph XML code using cognitive reasoning steps, avoiding model https://x.com/rohanpaul_ai/status/1914957917771718987
Tempus AI Signs Expanded Strategic Agreements with AstraZeneca and Pathos to Develop the Largest Multimodal Foundation Model in Oncology – Life Science Report https://lifesciencereport.com/news/nasdaq/tem/tempus-signs-expanded-strategic-agreements-with-astrazeneca-and-pathos-to-develop-the-largest-multimodal-foundation-model-in-oncology
