Self-interpretability: LLMs can describe complex internal processes that drive their decisions
Published on November 14, 2025 12:18 AM GMTThis post is a summary of our paper…
Recent Articles
View AllVCs abandon old rules for a ‘funky time’ of investing in AI startups
The goalposts are moving for AI startups when it comes to growth, product features, and…
Supervised fine-tuning as a method for training-based AI control
Published on November 13, 2025 10:25 PM GMTExecutive summaryThis post is a research update on…
Apple’s new App Review Guidelines clamp down on apps sharing personal data with ‘third-party AI’
Apple updates its App Store rules to restrict apps from sharing personal data with third-party…
Separating natural forests from other tree cover with AI for deforestation-free supply chains
Climate & Sustainability
OpenAI’s new LLM exposes the secrets of how AI really works
ChatGPT maker OpenAI has built an experimental large language model that is far easier to…
Google’s NotebookLM adds ‘Deep Research’ tool, support for more file types
The service rolling out "Deep Research," a tool that will automate and simplify complex online…
Google’s SIMA 2 agent uses Gemini to reason and act in virtual worlds
“SIMA 2 is a more general agent. It can complete complex tasks in previously unseen…
LinkedIn adds AI-powered search to help users find people
LinkedIn's AI-powered people search is rolling out to premium users in the U.S.
Google DeepMind is using Gemini to train agents inside Goat Simulator 3
Google DeepMind has built a new video-game-playing agent called SIMA 2 that can navigate and…