Bringing HDR video to Reels
Uploading HDR videos to Reels on Facebook and Instagram, with legible SDR UI elements and overlays.
Being first to market with rideshare on CarPlay and Android Auto
The blogpost discusses the cross-functional development process behind Lyft's launch of rideshare support for CarPlay and Android Auto, including partnerships with Apple and Google, design considerations for driver and rider safety, engineering challenges and iterations, user research and feedback, and future enhancements.
Developing reliable AI tools for healthcare
An AI system, CoDoC, is proposed in a joint paper with Google Research in Nature Medicine to develop reliable AI tools for healthcare, specifically to determine when to defer to a clinician for accurate interpretation of medical images.
A developer’s guide to prompt engineering and LLMs
How to effectively use prompt engineering to build LLM-based applications, with insights from GitHub's approach.
Snorkel Flow Summer 2023: faster, easier and more secure
Snorkel Flow Summer 2023: faster, easier, and more secure, with improved labeling functions, intelligent auto-suggest strategies, and streamlined annotation experiences.
Training Diffusion Models with <br> Reinforcement Learning
Exploring the intersection of training diffusion models with reinforcement learning
Snorkel Flow Summer 2023: faster, easier and more secure
Introducing Snorkel Flow Summer 2023, a release that improves the application creation process, shortens the iteration loop, and enhances security certifications.
A new way to look at data privacy
A privacy technique that maintains machine-learning model's performance while protecting sensitive data.
Study finds ChatGPT boosts worker productivity for some writing tasks
The report explores how generative AI, specifically ChatGPT, can enhance worker productivity for specific writing tasks.
AI helps household robots cut planning time in half
AI-powered PIGINet reduces planning time for household robots by leveraging machine learning to assess and filter feasible solutions in complex environments.
International ACM Conference on Research and Development in Information Retrieval (SIGIR) 2023
Apple's sponsorship and accepted paper presentation schedule at the International ACM Conference on Research and Development in Information Retrieval (SIGIR) 2023
Training Diffusion Models with Reinforcement Learning
The blogpost discusses training diffusion models with reinforcement learning, showing how they can be trained on downstream objectives directly using RL and introducing the Denoising Diffusion Policy Optimization (DDPO) algorithm.