Google DeepMindPredicting the past with Ithaca
Using Ithaca technology to predict and interpret damaged and ancient Greek inscriptions for a better understanding of past civilizations.
ChromiumA new speed milestone for Chrome
The blogpost discusses how Chrome achieved a new speed milestone in the M99 release, including the highest score in Apple's Speedometer browser responsiveness benchmark on Mac, performance improvements on Android, and optimizations such as ThinLTO, V8 Sparkplug compiler, short builtin calls, Freeze-Dried Tabs, and Isolated Splits.
Jane StreetResearch internships in our Tools and Compilers group
Announcement of research internships in the Tools and Compilers group
OpenAIEconomic impacts research at OpenAI
Exploring the economic impacts of AI through research at OpenAI
OpenAILessons learned on language model safety and misuse
Exploring insights and precautions for language model safety and usage
OpenAIA research agenda for assessing the economic impacts of code generation models
Exploring the economic impact assessment of code generation models in research agenda
Google DeepMindLearning Robust Real-Time Cultural Transmission without Human Data
Training agents for real-time cultural transmission without human data using deep reinforcement learning.
ChromiumChrome 100 Beta: Reduced User-Agent Strings, Multi-Screen Window Placement, and More
This technical blogpost discusses the new features and changes in Chrome 100 Beta, including reduced User-Agent Strings, Multi-Screen Window Placement, and more. It also covers origin trials, other features in this release, and deprecations and removals.
OpenAIA research agenda for assessing the economic impacts of code generation models
Exploring the framework for evaluating the financial effects of code generation models
OpenAILessons learned on language model safety and misuse
Exploring language model safety and misuse for AI developers.
Google DeepMindLearning Robust Real-Time Cultural Transmission without Human Data
Using deep reinforcement learning to enable artificial agents to learn and transfer cultural knowledge in real time.
Stanford AIUnderstanding Deep Learning Algorithms that Leverage Unlabeled Data, Part 1: Self-training
Understanding the self-training algorithm, a powerful paradigm for semi-supervised learning and domain adaptation, that leverages unlabeled data to improve accuracy.