They apparently had a temporary advantage in CoT reasoning. But all other models are on par now.
RESEARCH
-
Genie 2: Foundation World Model for Embodied Agent Training
By
–
Join Jack Parker-Holder & Philip Ball today at 5PM at the #ICLR2025 Google booth to discuss Genie 2, a foundation world model that can generate an endless variety of action-controllable, playable 3D environments for training & evaluating embodied agents. https://t.co/xPsrqa6j3s pic.twitter.com/5uioM3ejRy
— Google AI (@GoogleAI) 26 avril 2025Join Jack Parker-Holder & Philip Ball today at 5PM at the #ICLR2025 Google booth to discuss Genie 2, a foundation world model that can generate an endless variety of action-controllable, playable 3D environments for training & evaluating embodied agents. https://
deepmind.google/discover/blog/
genie-2-a-large-scale-foundation-world-model/
… -
Karl Popper AI and Automation of Scientific Discovery at ICLR2025
By
–
Love the analogy of The AI Scientist being an example of “Karl Popper AI” Happy that more people are excited about open-endedness, world models, and automation of discovery from @_rockt
’s fantastic talk at #ICLR2025 -
Open-Endedness and AI: Tim Rocktäschel’s ICLR2025 Keynote
By
–
Tim Rocktäschel’s keynote talk at #ICLR2025 about Open-Endedness and AI. “Almost no prerequisite to any major invention was invented with that invention in mind.” “Basically almost everybody in my lab at UCL and at DeepMind have read this book: Why Greatness Cannot Be Planned.”
-

Learning Algorithms Discover Randomization Benefits at ICLR2025
By
–
At 3PM today, Seijin Kobayashi & Johannes von Oswald will be at the #ICLR2025 Google booth to address the question, "Can learning algorithms discover when randomization is beneficial?" https://
arxiv.org/abs/2408.10818 -

Alpha ASI World Model Demo by Montreal AI
By
–
α-ASI World-Model Demo By http://
MONTREAL.AL : https://
github.com/MontrealAI/AGI
-Alpha-Agent-v0/blob/main/alpha_factory_v1/demos/alpha_asi_world_model/Alpha_ASI_World_Model.pdf
… #AGI #Agentic #AGIALPHA -

Google Gecko: Interpretable Metric for Image Prompt Adherence
By
–
At #ICLR2025? Stop by the Google booth today at 12PM to learn about Gecko, an interpretable, automatic metric for evaluating the prompt adherence of images & videos. Given an image/video + description, the model determines what is missing from the prompt. https://
arxiv.org/abs/2404.16820 -
UC Berkeley launches $12M grant program for AI startups
By
–
$12M grant for UC Berkeley-affiliated AI startups launches https://
dailycal.org/news/campus/re
search-and-ideas/12m-grant-for-uc-berkeley-affiliated-ai-startups-launches/article_0a18ce7c-591e-4208-b851-da7c011d5a21.html?utm_medium=social&utm_source=twitter&utm_campaign=user-share
… via @dailycal -

ProtNLM: Real-Time Protein Annotation with AI Models
By
–
Predicting protein annotations without a wet lab experiment is challenging. Today at 10AM, the #ICLR2025 Google booth will host a real-time demo of ProtNLM, a model developed together with EMBL-EBI to annotate proteins with short textual descriptions. https://
uniprot.org/help/ProtNLM -
Google AI Researcher Concludes Eight Year Montreal Lab Leadership
By
–
Today is my last day at Google. I started over 8 years ago, with a mandate to build a team doing bleeding edge AI research from Montreal, in what would be the first big tech AI research lab in the city. These years led to countless amazing scientific contributions from my team,